MCP Server

🔧 Fix MCP Endpoint Connectivity and Initialization Failures

🔄 Restoring Your ChatGPT MCP OAuth Connection

💡 What This Article Covers If you previously connected the Search Atlas MCP server to ChatGPT and the connection has stopped working — for example, ChatGPT no longer pulls your Search Atlas data, or the connector shows as disconnected — this guide explains why it happens and walks you through restoring it. 🔍 Why the Connection Drops The Search Atlas MCP integration uses OAuth 2.1 to link your ChatGPT account with your Search Atlas workspace. The connection can drop for a few common reasons: - Expired access token: OAuth tokens have a limited lifetime. Normally they refresh automatically, but occasionally the stored token becomes stale and ChatGPT stops recognizing the connection. - Revoked permissions: If access was revoked from your Search Atlas account or from your ChatGPT connector settings, the link breaks. - Workspace changes: Switching plans, changing account roles, or re-authenticating your Search Atlas login can invalidate the existing token. - ChatGPT connector cache: ChatGPT sometimes keeps a cached (stale) credential even after a token expires, which blocks reconnecting until the connector is fully removed and re-added. 🛠️ How to Reconnect in a Few Steps In most cases, fully removing and re-adding the connector fixes the problem: 1. Open ChatGPT and go to Settings → Connectors (sometimes labeled Apps). 2. Find Search Atlas in your list of connected connectors. 3. Open the options menu next to it and choose Remove or Disconnect. This clears the stale cached token. 4. Wait about a minute, then re-add the connector: search for Search Atlas in the connector directory, or re-enter the Search Atlas MCP server URL if you originally added it manually. 5. Complete the OAuth sign-in: you will be redirected to log in with your Search Atlas credentials and asked to authorize the connection. 6. Return to ChatGPT and confirm the connector shows as Connected. Important: Simply clicking Reconnect without first removing the connector often fails, because ChatGPT reuses the stale token. Always fully remove the connector first, then re-add it. ✅ Verifying the Connection Works After reconnecting, test it with a simple request such as: - "List the Search Atlas tools you can access." - "Pull my connected site's OTTO SEO overview." If ChatGPT responds with live data from your workspace, the connection is restored. ⚠️ Known Issue With a Fix in Progress We are aware of an issue where OAuth discovery endpoints return a 403 error on stale bearer tokens, which can prevent ChatGPT from reconnecting even after a fresh sign-in. A fix is being prepared for release. Until then, the full remove-and-re-add process above is the reliable workaround. If the 403 error appears immediately after re-adding the connector, wait 10–15 minutes and try once more. 🚧 If It Still Won't Connect Work through this quick checklist: - Confirm you are logging into the same Search Atlas account you originally connected. - Check that your Search Atlas subscription is active and that your role allows integrations. - Sign out of ChatGPT, clear your browser cache, sign back in, and retry. - Try the OAuth flow in a different browser or an incognito window to rule out cached cookies. - If ChatGPT reports a timeout right after authorization, wait a few minutes for the token to settle, then test again. 💬 Still Need Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔌 Restore Your ChatGPT MCP Connection

🔍 What This Article Covers If you previously connected Search Atlas to ChatGPT through MCP and the connection has stopped working — for example, ChatGPT no longer pulls your data or keeps asking you to sign in — this guide explains why it happens and how to restore the connection in a few minutes. ⚠️ Why the Connection Drops Search Atlas MCP uses OAuth 2.1 to authorize ChatGPT, and access tokens expire on a schedule for security. A dropped connection is usually temporary and easy to fix. Common causes include: - Expired token: The authorization token between ChatGPT and Search Atlas timed out and was not refreshed automatically. - Stale bearer token: A known issue can cause a 403 error on our discovery endpoints when an old token is cached. A permanent fix is rolling out now; until it reaches your account, reconnecting manually is the reliable workaround. - Disconnected app: The Search Atlas connector was removed in ChatGPT settings, either by you or by a workspace admin. - Account changes: A password reset, plan change, or workspace switch on the Search Atlas side can invalidate an existing session. 🛠️ How to Reconnect Step by Step The fastest fix is to fully remove the old connection and authorize it again. Follow these steps: 1. Open ChatGPT and go to Settings. 2. Open the Apps and Connectors section (name may vary slightly by ChatGPT version). 3. Find Search Atlas MCP in your list of connected apps and select Remove or Disconnect. This clears the stale token. 4. Re-add the connector using the same Search Atlas MCP listing or server URL you used originally. 5. Complete the sign-in screen with your Search Atlas account email and password — use the same account you use inside the Search Atlas platform. 6. Approve the permission prompts to finish authorizing the connection. 7. Return to your ChatGPT conversation and test the connection using the steps below. If ChatGPT shows a 403 error during the reconnect attempt, sign out of Search Atlas in your browser, sign back in, and repeat the steps above. This forces a fresh token to be issued. ✅ Confirm the Connection Works After reconnecting, verify everything is active: - In ChatGPT settings, the Search Atlas connector shows a Connected status. - Ask ChatGPT a simple question that requires your Search Atlas data, such as pulling details for one of your OTTO SEO sites or running a visibility check. - If the tool responds with live data from your account, the connection is restored. 💡 Tips to Avoid Future Disconnects - Reconnect promptly whenever ChatGPT shows a sign-in or authorization prompt — delaying can leave a stale token in place. - Use the same Search Atlas account in both the platform and ChatGPT to avoid session conflicts. - Keep an active Search Atlas subscription; a lapsed plan can interrupt the authorization. - After any password reset or workspace change, remove and re-add the connector to refresh the token. 🚀 Known Issue and Upcoming Fix Our engineering team is releasing a fix for the bug that causes 403 errors on stale bearer tokens during MCP reconnection. Once deployed, ChatGPT will refresh expired tokens automatically and manual reconnects should no longer be needed. Until the fix reaches your account, the remove-and-reconnect steps above will restore access every time. 🙋 Still Need Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔌 Official Search Atlas MCP Integrations

✅ Official repository ownership Search Atlas maintains an official MCP server repository, and only repositories published and maintained by Search Atlas itself should be treated as official. Community forks, look-alike repositories, and personal projects are not officially supported. To find the current official repository, refer to the setup materials published through official Search Atlas documentation or confirm with the Search Atlas support team before configuring anything. 🤖 Supported AI clients Which MCP clients are officially supported can change over time, so always verify support for your specific client in the official Search Atlas MCP setup materials. Before configuring a client such as Claude, ChatGPT, or Codex, confirm that official Search Atlas documentation covers that integration rather than relying on third-party tutorials or community configurations. 🛠️ Official setup resources Always use the README and documentation published by Search Atlas for installation instructions, required credentials, supported tools, and configuration examples. Avoid setup guides from unofficial sources. Before configuring an MCP client, confirm that you are following the current official instructions and using the correct Search Atlas credentials. Never paste credentials into unofficial configuration files or third-party services. 🔎 How to identify approved repositories 1. Confirm the repository is maintained under Search Atlas's official GitHub presence, not a personal account or an unrelated organization. 2. Check that the repository documentation identifies Search Atlas as the owner or maintainer. 3. Use setup instructions linked from official Search Atlas documentation or the repository itself. 4. Be cautious with repositories that use similar names, promise clients that official documentation does not mention, or request credentials through unrelated services. 5. When in doubt, treat a fork or personal repository as community-maintained unless Search Atlas explicitly identifies it as official. 6. If you are still unsure which repository is official, contact the Search Atlas support team to confirm before configuring anything. 💬 Need help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔑 Fix Search Atlas MCP ChatGPT Authentication Errors

🧭 Overview The Search Atlas MCP (Model Context Protocol) server lets you connect Search Atlas tools directly to AI assistants like ChatGPT. However, setting it up for the first time — or after an API key change — can produce authentication errors or a 'no actions available' message. This article walks you through the correct setup process and the most common fixes. ⚙️ How the Search Atlas MCP Works with ChatGPT Search Atlas uses OAuth 2.1 to authenticate your ChatGPT connection. When you link the MCP server, ChatGPT receives a signed JWT (JSON Web Token) that grants it access to your Search Atlas tools — including GSC data, keyword research, and LLM Visibility features. Important things to know before you begin: - The Search Atlas MCP server is a standard remote MCP server that works with any MCP client, ChatGPT included. Whether a given ChatGPT app or plan lets you add a custom MCP connector is decided on ChatGPT's side. - Your Search Atlas account must be on a plan that includes MCP/AI Agent access. - JWT tokens are tied to your current API key. If you rotate your API key or disconnect a site, you must re-authenticate — existing tokens are not automatically invalidated but will stop working correctly. 🚀 How to Install the Search Atlas MCP in ChatGPT 1. In your ChatGPT-connected environment (e.g., ChatGPT desktop app or a GPT Action setup), open the plugin or action configuration panel. 2. Select Add an action or Connect an MCP server, then paste in the Search Atlas MCP Server URL. 3. When prompted for authentication, choose OAuth login (an active subscription is required), or alternatively authenticate with a Search Atlas API key from Settings → API Keys passed via the X-API-Key header. 4. If using OAuth, ChatGPT will redirect you to a Search Atlas login screen — sign in with your account credentials to complete the OAuth handshake. 5. Once authorised, return to ChatGPT. Your Search Atlas tools should now appear as available actions. 🛠️ Fixing 'No Actions Available' Errors If you see a 'no actions available' message after connecting, work through these steps in order: 1. Confirm your ChatGPT app and plan support custom MCP connectors. Custom MCP connector availability depends on the customer's ChatGPT app and plan (ChatGPT's side). 2. Re-check your authentication. Confirm you completed OAuth login with an active subscription, or that the Search Atlas API key from Settings → API Keys you provided in the X-API-Key header is valid. 3. Complete the OAuth flow fully. The connection only activates after you sign in on the Search Atlas authorisation page. If you closed that window early, disconnect and reconnect the integration to restart the flow. 4. Check your plan permissions. MCP access is not available on all plans. Verify that your subscription includes AI Agent or MCP features under Avatar (top-right) → Billing (Plans & Top-ups). 5. Rotate your token after an API key change. If you recently rotated your Search Atlas API key or disconnected a site, your previous JWT may be stale. Disconnect the MCP integration in ChatGPT, then reconnect it using fresh credentials from Search Atlas. 6. Wait up to two minutes. After a successful OAuth handshake, ChatGPT may take a short moment to load available actions. Refresh or restart the session before concluding there is a problem. 🔐 Understanding Token Behaviour Search Atlas issues signed JWT tokens during the OAuth flow. Keep these points in mind to avoid future authentication issues: - Tokens are valid for a set period. If your session expires, you will need to re-authorise the connection. - Rotating your Search Atlas API key does not automatically revoke existing JWTs. However, those tokens will lose access to protected data. Always reconnect after a key rotation. - Disconnecting a site from Search Atlas does not instantly invalidate tokens linked to that site. Reconnect the MCP integration after any major account change to ensure your tokens are fresh and correctly scoped. ❓ Frequently Asked Questions - Can I use the Search Atlas MCP in the ChatGPT web app? The Search Atlas MCP server works with any MCP client, ChatGPT included. Whether you can add a custom MCP connector depends on your ChatGPT app and plan (ChatGPT's side). - How many tools does the MCP give me access to? When correctly authenticated, the MCP unlocks tools across GSC data, keyword research, and LLM Visibility — over 70 actions in total. - Do I need to reconnect if I change my Search Atlas password? Changing your password alone does not affect your MCP token, but if you reset your API key you should reconnect the integration. 💬 Still Need Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔌 Search Atlas MCP Client Compatibility Explained

🧭 Overview Search Atlas MCP (Model Context Protocol) lets AI clients access Search Atlas tools and data directly inside your AI workflow. Search Atlas MCP is built for any MCP client, ChatGPT included. It authenticates via OAuth login (requires an active Search Atlas subscription) or with a Search Atlas API key sent in the X-API-Key header (from Settings → API Keys). ✅ Supported MCP Clients Search Atlas MCP works with any MCP client, including: - Claude Desktop and Claude.ai - ChatGPT - Any other client that implements the standard MCP protocol with full tool-calling support If you are using one of these clients and experiencing a connection issue, that is a separate troubleshooting topic unrelated to client compatibility. 🔐 How to Authenticate Search Atlas MCP supports two authentication methods: - OAuth login — sign in with your Search Atlas account (requires an active Search Atlas subscription). - API key — send your Search Atlas API key in the X-API-Key header. You can generate one from Settings → API Keys. If you are on a supported client and still cannot connect, please refer to the Search Atlas MCP setup and troubleshooting guide. 💬 Need More Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🧪 Search Atlas MCP Server

The Search Atlas MCP Server (Model Context Protocol) allows AI tools to connect directly to Search Atlas and access its full suite of SEO, content, and marketing capabilities. Through MCP, your AI assistant can: - Access real data from your Search Atlas account - Execute actions across multiple products - Automate SEO workflows - Help diagnose issues using real platform signals MCP acts as the bridge between AI tools and Search Atlas, enabling faster, smarter, and more automated workflows. ⚙️ Available Services Search Atlas MCP provides access to a wide range of product areas: | | | | | --- | --- | --- | | Service | Status | Description | | OTTO SEO | ✅ Available | Full SEO automation, audits, fixes, indexing, link building | | Keywords | ✅ Available | Rank tracking, SERP analysis, competitor tracking | | Content Genius | ✅ Available | AI content creation, topical maps, grading, publishing | | Site Explorer | ✅ Available | Backlinks, keyword research, competitive analysis | | Local SEO | ✅ Available | GBP management, citations, heatmaps | | OTTO PPC | ✅ Available | Google Ads automation, campaign management | | Digital PR | ✅ Available | PR campaign workflows and opportunity discovery | | LLM Visibility | ✅ Available | AI/LLM brand monitoring | | Social Hub | ✅ Available | Social account management and scheduling | | LinkLab | ✅ Available | Link building marketplace | | Website Studio | ✅ Available | Website and landing page creation | | Press Release | ✅ Available | PR creation and distribution | | Cloud Stack | ✅ Available | Content distribution across platforms | | Brand Vault | ✅ Available | Brand assets and knowledge graph management | | Report Builder | 🔄 In Development | AI-powered reporting and summaries | 👉 Availability depends on your subscription and permissions. 🎯 What you can do with MCP With MCP, your AI tool can: 🔍 Analyze and audit - Run SEO audits - Review site health and performance - Identify issues and opportunities 🛠️ Optimize and execute - Generate SEO recommendations - Create and edit content - Apply fixes in bulk 📊 Manage workflows - Track projects and performance - Monitor keyword rankings - Manage campaigns and assets 🧠 Smarter answers with real data MCP enables AI assistants to respond using real account signals, not just generic responses. Example: “Why isn’t my rank tracker updating?” Your AI assistant can check: - tracking status - API/provider issues - quota limits - project configuration This allows faster and more accurate answers — often without needing support. 📚 Example Use Cases OTTO SEO - “Start a new project for example.com and run a full audit.” - “List my top projects by SEO health score.” - “Generate title tag suggestions for my top pages.” Keywords - “Show me ranking changes this week.” - “Add these keywords to tracking.” Content Genius - “Grade this article for SEO.” - “Regenerate this article in a more conversational tone.” Local SEO - “Show GBP heatmap performance.” - “Check citation status.” PPC - “Audit my Google Ads account.” - “Show wasted spend.” 🔌 Supported AI tools Search Atlas MCP can be used with: - Claude (browser & desktop) - Claude Code (CLI & desktop) - Cursor - VS Code - Warp - Custom applications via HTTP API 🔑 Getting Access To use MCP: 1. Open the Account Menu → Settings → API Keys 2. Copy your API key 3. Configure your AI client using the MCP endpoint 🌐 MCP Endpoints Full access https://mcp.searchatlas.com/mcp/ Namespace-specific access https://mcp.searchatlas.com/mcp/{namespace}/ Examples: - /mcp/otto/ - /mcp/keywords/ - /mcp/content-genius/ 👉 Namespace endpoints allow you to limit tools to a specific product. 🔐 Authentication Authentication is handled via API key: X-API-KEY: your_searchatlas_api_key_here Depending on the client, OAuth-based authentication may also be supported. 📈 Usage and Limits - No additional cost for MCP itself - Usage is tied to your Search Atlas plan - Tool limits (keywords, content, audits, etc.) still apply Rate limit - 40 requests per minute per user account - Exceeding this returns: - 429 Too Many Requests ⚠️ Important Notes 1. AI responses may not always be fully accurate — always review before applying changes 2. Some tools require confirmation before execution 3. Not all features may be available in every client 4. Tool availability depends on: - subscription - permissions - integration scope 🚧 Current Limitations While MCP provides powerful automation and diagnostics, keep in mind: - AI responses may not always surface all account details in a single response — breaking complex queries into smaller steps improves accuracy - Some tools require explicit user confirmation before executing destructive or financial actions (e.g., bulk deletions, credit purchases) - Not all tools are available in every namespace endpoint — use the full /mcp/ endpoint if a tool is missing from a namespace-specific one - Tool availability depends on your subscription, permissions, and integration scope 🌐 Community Edition (Testing) https://mcp.searchatlas.com/api/v1/mcp/community Key differences: - No authentication required - Strict rate limits - Limited tools - Not recommended for production ⚙️ Client Setup (Summary) Use the MCP endpoint below when connecting SearchAtlas to supported clients: URL:https://mcp.searchatlas.com/mcp/ Replace YOUR_API_KEY with the key copied from Account Menu → Settings → API Keys. 🤖 Claude (Browser & Desktop) Step 1 Go to Settings → Connectors → Add custom connector Step 2 Set the URL to: https://mcp.searchatlas.com/mcp/ Step 3 Add the X-API-KEY header with your API key Step 4 Save 💻 Claude Code (CLI) Run: claude mcp add --transport http searchatlas https://mcp.searchatlas.com/mcp/ --header "X-API-KEY: [REDACTED]" Then run claude in your terminal. 🖥️ Claude Code (Desktop) In Settings → Developer → Edit Config → claude_desktop_config.json: { "mcpServers": { "searchatlas": { "command": "npx", "args": [ "mcp-remote", "https://mcp.searchatlas.com/mcp/", "--header", "X-API-KEY: ${X_API_KEY}" ], "env": { "X_API_KEY": "[REDACTED]" } } } } 🧭 Cursor In Settings → Cursor Settings → MCP & Integrations: { "mcpServers": { "searchatlas": { "url": "https://mcp.searchatlas.com/mcp/", "headers": { "X-API-KEY": "[REDACTED]" } } } } 🧩 VS Code Create a .vscode/mcp.json file in your workspace: { "mcpServers": { "searchatlas": { "url": "https://mcp.searchatlas.com/mcp/", "headers": { "X-API-KEY": "[REDACTED]" } } } } 🚀 Warp In Settings → AI → Manage MCP Servers → Add: { "searchatlas": { "command": "npx", "args": [ "mcp-remote", "https://mcp.searchatlas.com/mcp/", "--header", "X-API-KEY: ${X_API_KEY}" ], "env": { "X_API_KEY": "[REDACTED]" }, "working_directory": null } } This setup allows supported clients to connect to the SearchAtlas MCP endpoint using your API key. Make sure you replace the placeholder value with the key copied from Account Menu → Settings → API Keys before saving your configuration. 🔧 Direct API Integration List tools curl -X POST "https://mcp.searchatlas.com/mcp/" \ -H "Content-Type: application/json" \ -H "X-API-KEY: YOUR_API_KEY" \ -d '{"jsonrpc":"2.0","method":"tools/list","id":1}' Call a tool curl -X POST "https://mcp.searchatlas.com/mcp/" \ -H "Content-Type: application/json" \ -H "X-API-KEY: YOUR_API_KEY" \ -d '{ "jsonrpc":"2.0", "method":"tools/call", "params":{"name":"otto_find_project_by_hostname","arguments":{"hostname":"example.com"}}, "id":1 }' 🧠 Troubleshooting Authentication issues - Check API key - Confirm header format (X-API-KEY) - Verify subscription Connection issues - Confirm endpoint: ​https://mcp.searchatlas.com/mcp/ - Check client configuration Rate limit errors - Wait and retry - Avoid repeated rapid calls Missing tools - Check subscription access - Verify permissions - Confirm namespace Common errors | | | | | --- | --- | --- | | Code | Meaning | Action | | 401 | Unauthorized | Check API key | | 403 | Forbidden | Check permissions | | 429 | Rate limit | Wait and retry | | -32601 | Method not found | Check tool name | | -32602 | Invalid params | Verify inputs | 💡 Pro Tip Use MCP together with the Search Atlas dashboard to maintain full visibility over: - performance - reports - live results The Search Atlas MCP Server connects AI tools directly to Search Atlas. It enables you to: - automate SEO, content, and marketing workflows - access real account data through AI - execute actions across multiple product areas - work faster without switching platforms MCP is the foundation that connects AI-driven workflows with real Search Atlas capabilities.

🔌 How to Connect Claude to Search Atlas via MCP

🧩 What Is MCP and Why Does It Matter? The Model Context Protocol (MCP) is an open standard that allows AI assistants like Claude to connect directly to external tools and data sources. By connecting Claude to Search Atlas via MCP, you can bring your SEO data, keyword insights, and site metrics directly into your Claude conversations — enabling AI-assisted workflows without switching between platforms. This integration allows Claude to query Search Atlas data on your behalf, helping you analyze keywords, review site metrics, and surface actionable SEO insights through natural language prompts. ✅ Prerequisites Before you begin, make sure you have the following in place: - An active Search Atlas account with API access enabled - Claude for Desktop installed on your machine (available from Anthropic) - Your Search Atlas API key (retrievable from Account Menu → Settings → API Keys) - Basic familiarity with editing configuration files on your operating system 🔑 Step 1: Retrieve Your Search Atlas API Key You will need your API key to authenticate the MCP connection between Claude and Search Atlas. 1. Log in to your Search Atlas account. 2. Navigate to Account Menu → Settings → API Keys. 3. Copy your key to a safe location. 4. Keep this key private — do not share it publicly or commit it to version control. ⚙️ Step 2: Configure Claude for Desktop to Use the Search Atlas MCP Server Claude for Desktop uses a configuration file to register MCP servers. Follow the steps below to add the Search Atlas MCP server. 1. Open Claude for Desktop on your computer. 2. Open the Claude for Desktop configuration file in a text editor. Refer to the official Anthropic Claude for Desktop documentation for the exact location of this file on your operating system. 3. Add the Search Atlas MCP server entry to the configuration file, providing your API key as specified in the Search Atlas MCP documentation. 4. Save the configuration file. 5. Restart Claude for Desktop to apply the changes. Refer to the official Search Atlas MCP documentation for the exact server name, command, and parameter fields required in your configuration file, as these may be updated over time. 🧪 Step 3: Verify the Connection Once Claude for Desktop has restarted, confirm that the integration is working correctly. 1. Open a new conversation in Claude for Desktop. 2. Check that Search Atlas tools are available. Refer to the Anthropic Claude for Desktop documentation for guidance on how loaded MCP tools are displayed in your version of the application. 3. Type a test prompt referencing your Search Atlas data to confirm the connection is active and returning results. 🛠️ Troubleshooting If the integration does not appear to be working after following the steps above, consider the following: - Double-check that your Search Atlas API key is entered correctly in the configuration file with no extra spaces or characters. - Ensure your Search Atlas account has API access enabled. - Confirm you are using a version of Claude for Desktop that supports MCP integrations. - Review both the Search Atlas MCP documentation and the Anthropic Claude for Desktop documentation for any version-specific setup requirements. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Integrations & External Systems

The MCP allows your AI assistant to trigger actions that connect Search Atlas workflows with external destinations such as CMS platforms, publishing systems, and other integrated services. This article explains what MCP integrations are, how they work, what types of actions are supported, and how to safely execute workflows that involve external systems. 🧠 What Integrations Mean in MCP In the MCP context, integrations refer to actions that: - send data outside of Search Atlas - publish or distribute content - connect with third-party platforms - trigger external workflows These are not separate tools — they are extensions of existing Search Atlas product areas (such as Content, GBP, PPC, PR) that interact with external systems. 🔍 What You Can Do Using MCP with integrations, you can: - publish content to CMS platforms (e.g., WordPress) - distribute press releases - submit data to external services (indexing, citations) - execute outreach or PR workflows - trigger actions that interact with connected accounts These actions extend beyond internal data and result in real-world changes or external outputs. About "connected accounts": Before MCP can execute any integration action, the relevant external system must be connected to your Search Atlas workspace. This typically means configuring a CMS connector (for WordPress and other CMS platforms), connecting your Google Business Profile, or authorizing the third-party service you want to publish to. CMS connections are set up via the CMS connector inside Search Atlas — without an active connector, MCP publish actions will fail even if your AI assistant successfully prepares the request. If you are unsure whether your destination is connected, check the CMS connector / integrations area in the Search Atlas dashboard before requesting a publish action. ⚙️ Step-by-Step: Execute an Integration Action Step 1 Open your MCP-connected AI assistant Step 2 Ask for an action that involves an external system Examples: - “Publish this article to WordPress” - “Submit these URLs for indexing” - “Distribute this press release” Step 3 The assistant prepares the integration action At this stage, it determines: - the destination system - the required data - whether approval is needed before execution If approval is required, the assistant will surface the prepared action for your review and wait for your confirmation before proceeding. Step 4 Review the action details This is important because integration actions may: - publish publicly - trigger external processes - consume credits or HDC Step 5 Approve the action (if required) Actions that involve: - external publishing - spending - irreversible effects may require confirmation before execution. Step 6 The MCP executes the integration The action is sent to the external system using your connected account and permissions. After execution, verify the outcome in the destination system (for example, check that the post appears in your CMS, that the press release shows as distributed, or that submitted URLs appear in their respective indexing queues). If the action does not appear as expected, see the Troubleshooting section below. 🌐 Types of Integration Actions 📰 Publishing - CMS publishing (e.g., WordPress) - scheduled content distribution Note on supported publishing paths: WordPress and other CMS publishing is supported via the Search Atlas CMS connector. Content Genius posts can be published to Website Studio sites, as well as to WordPress and through Universal CMS connectors. Please distinguish between supported CMS connectors (such as WordPress via the CMS connector) and integration paths that are still in development. 📡 Submissions - URL indexing - citation submissions - directory distribution 📣 Outreach & PR - press release distribution - outreach campaigns Note on press release indexation: Press release URLs distributed through Signal Genesys are now subject to an automated daily indexation check. After distribution, the system will periodically verify whether the released URLs have been indexed by search engines, and you can expect to see indexation status reflected in the corresponding press release report over time without needing to manually re-check each URL. 🔗 External Execution - actions that trigger systems outside Search Atlas infrastructure 📍 GBP (Google Business Profile) - GBP-related MCP actions (posts, updates, profile interactions) Note on GBP MCP responses: GBP MCP now returns responses using structured output. This means GBP-related responses are returned as predictable, schema-shaped objects rather than free-form text. If your AI assistant or downstream automation was previously parsing GBP MCP responses as plain strings, you may need to update that parsing to read the structured fields directly — most MCP-aware assistants handle this automatically. All of these are part of the broader MCP toolset interacting with external services. ⚠️ Important Behavior Differences Integration actions are different from internal actions in key ways: 1. They may be irreversible Once content is published or submitted, it may not be easily undone 2. They may require approval The MCP may pause execution and request confirmation 3. They may consume credits Depending on the action, they may use: - Search Atlas quota - Hyperdrive Credits (for premium distribution) 4. They depend on external systems Execution success may depend on: - third-party platform availability - account connection status - external API behavior 5. They may depend on plan tier or backend enablement Not every MCP endpoint is enabled for every plan or for general availability. Some endpoints (for example, certain bulk SERP endpoints such as se_bulk_serps) require explicit backend enablement and are not opened up to all accounts by default, even when your quota dashboard shows available credits. If you call one of these endpoints and receive an ACCESS_DENIED response despite having sufficient quota, this typically indicates a backend access restriction rather than a quota problem — please contact Search Atlas support so the endpoint can be reviewed for your account. 🔐 Step-by-Step: What to Check Before Running Integration Actions Step 1 Confirm the destination system is connected Example: CMS, GBP, or external platform Step 2 Check your quota or HDC balance Step 3 Understand whether the action is public or external Step 4 Review the action carefully Step 5 Approve only if you are ready to execute 📋 Plan Availability & Endpoint Access MCP integration actions are not uniformly available across all Search Atlas plans. The general guidance is: - Core publishing & submissions (CMS publish via supported connectors, URL indexing, citations): generally available to plans that include the relevant product area (Content, OTTO, Local). - Press release distribution via Signal Genesys: requires Hyperdrive Credits (HDC) and the PR distribution add-on. - GBP MCP actions: require a connected Google Business Profile on a plan that includes Local features. - Bulk SERP and other advanced/data-volume endpoints (e.g., se_bulk_serps): may require special enablement on your account even if your quota dashboard shows availability. These are not enabled by default for general MCP usage. If you are unsure whether a specific MCP endpoint is included in your plan — or if you receive an ACCESS_DENIED response despite a healthy quota balance — contact Search Atlas support with the endpoint name and the request your assistant attempted. Support can confirm whether the endpoint requires backend enablement for your workspace. 🛠️ Troubleshooting Integration Issues Issue 1 — Action fails after approval Possible causes: - external system not connected - authentication issue - API error on the third-party platform Issue 2 — Action does not execute Possible causes: - waiting for approval - insufficient quota - permission restrictions Issue 3 — Unexpected output Possible causes: - incorrect input data - incomplete instructions - mismatch between requested action and integration behavior Issue 4 — "Website not connected" error on publish, even though the site appears connected If a publishing action (for example, publishing to WordPress) fails with a "website not connected" error while the site clearly shows as connected in your Search Atlas dashboard, this is a known state mismatch between the dashboard view and the CMS connector / Atlas Brain status used by MCP. Try the following, in order: - Open the CMS connector / Atlas Brain integration for that site and confirm its status there (not just the project-level dashboard view). Reconnect or re-authorize the connector if its status is not "active". - Attempt a manual publish from the relevant Search Atlas tool (e.g., Content Genius) as a fallback. If the manual publish succeeds, the connector itself is healthy and the issue is isolated to the MCP path. - If the error persists after reconnecting and the manual publish also fails, contact Search Atlas support with the site URL, the connector status you see, and the exact error message — this is the information support needs to investigate the connection state on the backend. Issue 5 — ACCESS_DENIED with sufficient quota If an MCP call returns ACCESS_DENIED even though your quota dashboard shows enough credits, the endpoint may not be enabled for your account or plan. This is most often seen on advanced data endpoints (for example, bulk SERP endpoints). Contact support so the endpoint can be reviewed for backend enablement on your workspace. 🔁 How Integrations Fit into MCP Workflows Integration actions are typically the final step in a workflow: 1. Generate or prepare data (content, audit, campaign) 2. Review and refine 3. Execute external action (publish, submit, distribute) This ensures that external actions only happen after validation and approval. 🧠 Best Practices Review before executing Always verify content or data before publishing or submitting Use approval as a checkpoint Approval prompts are designed to prevent unintended actions Separate preparation from execution Generate and review first, then execute integrations Be aware of external impact Integration actions often have real-world visibility MCP integrations allow your AI assistant to move beyond internal workflows and interact with real-world systems. By combining preparation, review, and controlled execution, you can safely publish, distribute, and trigger external actions using the same Search Atlas capabilities, permissions, and quota rules tied to your account.

📘 MCP Rate Limits & API Limits

The Search Atlas MCP follows the same rate limits and system constraints as the underlying Search Atlas API. This means all requests made through MCP are subject to standard API limits, including request throttling and usage constraints. 🧠 How Rate Limits Work MCP does not introduce its own rate limiting system. Instead, it relies on the existing Search Atlas API limits. 👉 Every request made through your AI assistant is processed as an API call. 👉 If too many requests are made in a short period of time, the system may temporarily block additional requests. You can view your current quota and usage status from Avatar → Billing → Plans & Top-ups (Total Quota) in your Search Atlas dashboard. For specific rate and quota thresholds applicable to your plan, refer to the Search Atlas API limits reference documentation. ⚙️ Step-by-Step: What Happens When You Hit a Rate Limit Step 1 You send multiple requests through your AI assistant Example: - Running multiple audits - Generating large batches of content - Executing repeated tool calls Step 2 The MCP forwards requests to the Search Atlas API Step 3 The API evaluates request volume Step 4 If the limit is exceeded: - The request is rejected - A rate limit error is returned (commonly a 429 error) - A 429 error can result from two distinct causes: (1) request-frequency rate limiting, or (2) plan quota exhaustion. Check your quota dashboard when you receive a 429 to determine which condition applies. 🚦 429 vs. ACCESS_DENIED: Understanding the Difference A 429 error and an ACCESS_DENIED error are not the same and should be troubleshooted differently: - 429 (rate limit / quota exhausted): Returned when your request frequency is too high or your plan quota has been exhausted. - ACCESS_DENIED: Returned when the specific MCP endpoint you are calling is not publicly enabled for your account. Some MCP endpoints are not yet publicly available for general use, so ACCESS_DENIED does not necessarily mean you have exceeded your plan quota. If you see ACCESS_DENIED despite your dashboard showing available quota, the endpoint you are calling may not be publicly enabled for MCP. Contact support to confirm whether the specific MCP tool is available on your plan. ⚠️ Common Scenarios That Trigger Rate Limits 1. Rapid repeated requests 2. Large batch operations executed too quickly 3. High-frequency automation tasks ⚙️ Step-by-Step: How to Resolve Rate Limit Errors Step 1 Stop sending new requests Step 2 Wait at least 60 seconds before retrying. For persistent 429 errors related to quota exhaustion, the limit typically resets at the start of your next billing cycle. Step 3 Retry the request Step 4 If needed, reduce request frequency 🛠️ Unexpected Quota or Limit Issues If you believe your quota usage is being over-reported or you are hitting limits unexpectedly early in a billing cycle, contact support — in some cases backend counter drift or billing-sync defects can cause incorrect quota readings that require an engineer to correct. Additionally, if your account shows a limit of 0 or is unexpectedly fully blocked despite low actual usage, this may be a plan/quota calculation defect. In that case, contact support rather than simply waiting or reducing request frequency. 🔁 Best Practices To avoid hitting rate limits: - Space out large operations - Avoid sending repeated requests in rapid succession - Break large workflows into smaller steps ⚠️ Important Notes 1. Rate limits apply equally to MCP and direct API usage 2. Limits are enforced automatically by the system 3. MCP does not override or bypass these limitsRate limits apply equally to MCP and direct API usage 4. Limits are enforced automatically by the system 5. MCP does not override or bypass these limits MCP uses the same infrastructure and protections as the Search Atlas API, including rate limits. By pacing your requests and structuring workflows efficiently, you can avoid interruptions and maintain smooth execution when working through your AI assistant.

🛠️ Fix WebSocket 403 Errors on MCP Bulk Send

🔍 Overview When using the Search Atlas AI Agent via the Model Context Protocol (MCP), calls to ws_bulk_send_message may return an HTTP 403 Forbidden error on the WebSocket channel. This article explains why this happens, how to confirm the root cause, and what steps to take to restore normal functionality. ⚠️ What Causes the 403 Error? An HTTP 403 (Forbidden) response on a WebSocket upgrade or message-send request means the server received your connection attempt but rejected it due to a missing, expired, or malformed JSON Web Token (JWT). Because ws_bulk_send_message relies on the same authenticated WebSocket channel used by Website Studio's MCP tools, a broken token blocks every bulk message call — even when the underlying message delivery actually succeeds on the server side. A 403 error on ws_bulk_send_message is almost always caused by one of the following: - Expired or invalid JWT token: The authentication token used to open the WebSocket connection has expired or was not issued correctly, causing the server to reject subsequent message calls. - MCP bulk sends disabled at the project level: In some cases, the ability to send bulk messages via MCP can be turned off for a specific project, either by a configuration change or a platform update. - Incorrect WebSocket endpoint: An older integration may be pointing to a deprecated REST endpoint instead of the correct WebSocket channel, resulting in auth failures. An earlier version of Website Studio's V2 integration attempted REST endpoint calls instead of routing traffic through the correct WebSocket channel, producing 403 and 404 responses. - False failure reporting: In some resolved cases, ws_bulk_send_message reported a 403 failure even though the message was delivered successfully. This was a known bug that has since been fixed. - create_project dependency: Because create_project and ws_bulk_send_message share the same authenticated session setup, a broken project-creation flow could leave the WebSocket channel in an unauthenticated state for all subsequent calls. Common symptoms include: - All ws_bulk_send_message calls return 403 immediately without queuing messages. - Website Studio editing tools (create_project, page edits) also fail or return errors when accessed via MCP. - The AI Agent log shows JWT auth errors alongside the 403 response. - Some messages appear delivered in the UI despite the 403 error in the API response — this is a known false-failure behaviour that has been patched. ✅ How to Confirm the Root Cause Before taking corrective action, follow these steps to identify exactly what is causing the error in your project: 1. Check your JWT token validity. Inspect the token being passed in your MCP connection headers. Tokens typically expire after a set period. If the token is expired, re-authenticate to generate a new one and re-establish the WebSocket connection. 2. Verify your WebSocket endpoint. Confirm that your integration is connecting to the correct WebSocket URL for the Search Atlas agent channel. If you see REST-style URLs (e.g., /api/v2/...) in your request logs, your integration may be misconfigured and needs to be updated to use the WebSocket endpoint. 3. Test with a fresh project. Create a new project and attempt a ws_bulk_send_message call. If the call succeeds on a new project but fails on the original one, the issue is likely tied to that specific project's configuration. 4. Check whether MCP sends are enabled. In your project settings, confirm that MCP-based messaging is active. If the option appears greyed out or unavailable, the feature may have been disabled for your account or project. Note that a true MCP-disabled state returns a configuration-level error message (such as "MCP tool not available" or "feature disabled") indicating the feature is turned off — it does not return an HTTP 403. If you are receiving a 403 specifically, MCP messaging is enabled but the request is being rejected at the authentication layer. 5. Review recent platform updates. If the error appeared suddenly without any changes on your end, it may be related to a recent platform update. Check the Search Atlas changelog or status page for any relevant notices. 🔧 Steps to Restore ws_bulk_send_message Once you have identified the cause, use the appropriate fix below: - Expired JWT token: Re-authenticate your MCP client using your Search Atlas credentials to obtain a fresh JWT token. In Search Atlas, go to Account Menu → Settings → API Keys and regenerate your API key. Copy the new key and update it in your MCP client configuration. Restart your AI Agent or MCP client to force a fresh WebSocket handshake with the new JWT. Do not reuse a cached session token from before a fix was deployed. Update your WebSocket connection to use the new token before retrying ws_bulk_send_message. - Wrong endpoint: Update your integration to point to the correct WebSocket channel. Remove any references to legacy REST endpoints. Contact support via the chat widget if you need the current endpoint details for your account. - Project-level configuration issue: If a specific project appears to have MCP messaging disabled or misconfigured, try removing and re-adding the project connection within the AI Agent settings. This resets the channel configuration. Also confirm that the target project was created successfully by opening Website Studio and verifying it exists. If create_project previously failed silently, the project may be missing, which will cause all subsequent ws_bulk_send_message calls to 403 because there is no valid channel to attach to. Re-run create_project if needed. - False failure on successful delivery: If your messages are actually being delivered despite the 403 error appearing in your logs, this was a known reporting bug that has been resolved in recent platform updates. Ensure your platform is on the latest version and verify message delivery directly in Website Studio before assuming a failure occurred. If you retried sends during the false-failure window, your project may contain duplicate content — review the affected pages in Website Studio and remove any duplicates manually. - Test with a single message first. Before running a bulk send, call ws_bulk_send_message with a single small payload to confirm the channel is authenticated and returning 200. 🚫 How to Confirm if MCP Sends Are Disabled To check whether MCP bulk messaging has been disabled for your project: 1. Click Atlas Agent in the header (e.g., 'Launch Atlas Agent' on Home). 2. Navigate to the project in question and open its MCP Channel Settings. 3. Look for a toggle or status indicator labelled Bulk Messaging or MCP Sends. If it is off, enable it and save your changes. 4. Retry the ws_bulk_send_message call to confirm the error is resolved. If the toggle is not visible or cannot be enabled, the feature may require reactivation at the account level by a member of our team. If you see a message such as "MCP tool not available" or "feature disabled", contact support to confirm your plan includes MCP access. 💡 Prevention Tips - Always handle JWT token expiry gracefully in your MCP integration by implementing automatic re-authentication before the token expires. - Log both the HTTP status code and the actual delivery outcome separately so you can distinguish between true failures and false reporting errors. - After any Search Atlas platform update, do a quick test call on your most active projects to confirm WebSocket connectivity is intact. - After any bulk send retry cycle, review affected pages in Website Studio to check for duplicate content that may have been created during a false-failure window. 🙋 Still Need Help? If you have followed the steps above and are still experiencing 403 errors on ws_bulk_send_message, our team can review your account's WebSocket logs directly. To reach us, open the chat widget in the bottom-right corner of the platform and type human teammate.

📘 Understanding MCP Cost Before Running Actions

Before executing actions through MCP, you can check whether they will consume Search Atlas quota to avoid unexpected usage. This allows you to plan your actions and ensure you have enough credits before running large or resource-intensive tasks. 🧠 Why This Matters Some MCP actions: - consume quota (when work is produced) - require confirmation (for paid or irreversible actions) Checking cost beforehand helps you: - avoid failed executions - control usage - plan workflows more efficiently ⚙️ Step-by-Step: Check Cost Before Running an Action 📌 Step 1 Open your MCP-connected AI assistant 📌 Step 2 Ask about the action before executing it Examples: - “Will this action consume credits?” - “How much quota will this use?” - “Check cost before running this task” 📌 Step 3 The assistant evaluates the request It determines whether the action: - is read-only - produces work - requires quota or Hyperdrive Credits 📌 Step 4 Review the response You will be informed if the action: - consumes quota - requires approval - is free (read-only) 🔍 How MCP Determines Cost MCP follows the core rule: 👉 Quota is charged when work is produced This means: - Generating content → consumes quota - Running audits → consumes quota - Submitting data → consumes quota - Viewing data → does NOT consume quota 🔐 Approval-Based Actions Some actions require confirmation before execution, especially when they: - consume Hyperdrive Credits - trigger external systems - are irreversible or visible externally In these cases: - The assistant prepares the action - You are prompted to confirm - Execution only proceeds after approval ⚠️ Important Notes - Checking cost is a read-only action - It does NOT consume Search Atlas quota - It only uses AI assistant tokens 🔁 Best Practice Before running large workflows (audits, campaigns, content generation), always check: - your available quota - whether the action will consume credits This helps avoid interruptions and ensures smoother execution. MCP allows you to understand costs before executing actions, giving you full visibility and control over your usage. By checking costs in advance, you can confidently run tasks knowing exactly how they will impact your Search Atlas credits. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for SEO Research & Backlinks

Use MCP to analyze backlinks, research competitors, and identify SEO opportunities using the Search Atlas research tools. Through MCP, your AI assistant can access backlink data, keyword insights, and competitor analysis to support your SEO strategy. 🧠 What You Can Do Using MCP for SEO research, you can: - Analyze backlinks and referring domains - Research competitors - Identify keyword opportunities - Discover content gaps - Evaluate domain authority and link profiles ⚙️ Analyze Backlinks 1. Open your MCP-connected AI assistant. 2. Ask for backlink data, for example: "Show me backlinks for my domain" or "Analyze my backlink profile." 3. The assistant retrieves backlink data. 4. Review the results, including referring domains, link sources, and backlink distribution. 🔍 Research Competitors 1. Ask your assistant, for example: "Analyze my competitors" or "Show me competitor backlink profiles." 2. The assistant gathers competitor data. 3. Review insights such as competitor domains, backlink comparisons, and SEO positioning. 🔑 Identify Keyword Opportunities 1. Ask your assistant, for example: "Find keyword opportunities for my site" or "Show keyword gaps between me and competitors." 2. The assistant analyzes keyword data. 3. Review the recommendations. 🧩 Discover Content Gaps 1. Ask your assistant, for example: "Find content gaps in my niche" or "What topics am I missing?" 2. The assistant compares your content against competitors. 3. Review the suggested topics and gaps. ⚠️ Important Notes - Data retrieval and analysis actions may consume Search Atlas quota. - Results depend on available data within the platform. - All actions follow your plan entitlements and permissions. 🔁 How This Fits Together The SEO research workflow through MCP typically follows: analyze your backlink profile, research competitors, identify keyword gaps, then discover content opportunities. All steps are executed through your AI assistant using the same tools available in the Search Atlas platform. MCP allows you to perform deep SEO research through conversation. From backlink analysis to competitor insights and keyword discovery, your AI assistant can execute each step using the same capabilities, permissions, and quotas available in your Search Atlas account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Social Media & PR

Use MCP to create, manage, and distribute social media content and digital PR campaigns using the Search Atlas Social and PR tools. Through MCP, your AI assistant can generate posts, manage publishing workflows, and execute outreach campaigns — all through natural language. 🧠 What You Can Do Using MCP for Social Media & PR, you can: - Create and publish social media posts - Manage social content workflows - Launch outreach campaigns - Generate press releases - Execute digital PR strategies ⚙️ Create Social Media Posts 1. Open your MCP-connected AI assistant. 2. Ask it to create social content, for example: "Create a social media post for my business" or "Write posts for my campaign." 3. The assistant generates post content. 4. Review the content. 5. Approve for publishing, if required. 📤 Publish Social Content 1. Ask your assistant to publish or schedule the post, for example: "Publish this post to my social channels" or "Schedule this content." 2. The assistant prepares the publishing action. 3. Confirm the action, if required. 4. The content is published or scheduled. 🧩 Launch Outreach Campaigns 1. Ask your assistant to start an outreach or PR campaign, for example: "Start an outreach campaign" or "Create a PR campaign." 2. The assistant prepares the campaign. 3. Review the campaign details. 4. Approve execution. 📰 Generate Press Releases 1. Ask your assistant, for example: "Write a press release for my business." 2. The assistant generates the press release. 3. Review and edit the content. 4. Approve distribution, if required. ⚠️ Important Notes 1. Content generation and campaign creation may consume Search Atlas quota. 2. Publishing and scheduling actions may require confirmation. 3. All actions follow your plan entitlements and permissions. 🔁 How This Fits Together The Social Media & PR workflow through MCP typically follows: create content, review and refine, publish or schedule, then launch outreach or PR campaigns. All steps are executed through your AI assistant using the same tools available in the Search Atlas platform. MCP allows you to manage your social media and PR workflows through conversation. From creating posts to launching campaigns and distributing press releases, your AI assistant can execute each step using the same capabilities, permissions, and quotas available in your Search Atlas account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 How to Connect MCP (Overview)

There is a single canonical connection target and a single authentication model for the Search Atlas MCP. Everything else — which client, which plan, or which namespace — is a variation on these two basics. This article explains the connection endpoint, the recommended authentication method, and what to check if the connection fails. 🌐 Connection Endpoint The public MCP endpoint serves the full catalog of Search Atlas tools: https://mcp.searchatlas.com/mcp/ The server also exposes per-product namespaces, for example /mcp/otto/ and /mcp/local-seo-gbp/, but the root endpoint aggregates all namespaces and is what customers should use unless there is a specific reason to scope access. Common tools such as brand vault, knowledge graph, and billing are available in every namespace. 🔐 Authentication The MCP server validates credentials issued by the main Search Atlas platform. It does not issue its own credentials. OAuth is the recommended and supported path for all customer-facing scenarios. The server also accepts a Search Atlas API key via the X-API-Key header, but that path exists for scripts and server-to-server integrations where an interactive browser login is not possible. It is not recommended for regular customer use. If a customer asks which to use, steer them to OAuth. ⚙️ Connect MCP 1. Use the main endpoint: https://mcp.searchatlas.com/mcp/ 2. Add that endpoint in your MCP-compatible client, such as Claude Desktop, Claude.ai, Cursor, or ChatGPT connectors. 3. When the client starts the connection flow, complete the one-time Search Atlas login in the browser. This is the OAuth flow recommended and supported for customer-facing scenarios. 4. After you sign in, the client holds and refreshes the credential automatically. From that point forward, the MCP connection remains saved in the client. ⚠️ If the Connection Fails The three most common causes, in order: 1. You are not already logged into Search Atlas when the browser window opens. Sign in to dashboard.searchatlas.com first, then try again. 2. Your plan does not include MCP access. Check your plan entitlements — MCP access is controlled at the plan level. 3. Your session is stale or expired. Disconnect the MCP in your client and reconnect to trigger a fresh login. To connect MCP successfully, use the root Search Atlas endpoint and the OAuth login flow. Once connected, your client can access the Search Atlas tool catalog available under your account, subject to your plan entitlements and quota rules. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Brand Vault & Knowledge Graph

Use MCP to manage your Brand Vault and Knowledge Graph, which store your business data, brand voice, and structured information used across Search Atlas tools. The Brand Vault acts as a centralized data layer that powers content generation, SEO, and automation workflows. Through MCP, your AI assistant can read and update this data directly. 🧠 What You Can Do Using MCP with Brand Vault and the Knowledge Graph, you can: - Store and manage business information - Define brand voice and tone - Provide source material for AI content generation - Maintain structured business data used across tools ⚙️ View Your Brand Data 1. Open your MCP-connected AI assistant. 2. Ask to retrieve your Brand Vault data, for example: "Show my brand information" or "What data is stored in my Brand Vault?" 3. The assistant retrieves your stored data, which may include business details, brand voice, and structured information used across the platform. ✍️ Update Brand Information 1. Ask your assistant, for example: "Update my business description in Brand Vault" or "Change my brand tone to more professional." 2. The assistant prepares the update. 3. Review the changes. 4. Confirm, if required. 5. The MCP updates your Brand Vault. 🧩 Use Brand Data in Content 1. Ask your assistant, for example: "Write content using my brand voice." 2. The assistant retrieves Brand Vault data. 3. Content is generated using your tone, your business information, and your stored sources. 🧱 How the Knowledge Graph Works The Knowledge Graph is the structured representation of your business data inside Search Atlas. It includes business identity, services and offerings, structured attributes, and supporting content. This data is used across content generation, SEO tools, and automation workflows. ⚠️ Important Notes - Updating Brand Vault data affects multiple product areas. - Content generation relies on this data for accuracy. - Changes should reflect real business information. - Some updates may require confirmation before execution. 🔁 How This Fits Together The Brand Vault workflow through MCP typically follows: store business data, define brand voice, use that data in content generation, then update it as your business evolves. All actions are executed through your AI assistant using the same system as the Search Atlas platform. 🧠 Best Practice Keep your Brand Vault accurate and complete. This ensures better AI-generated content, more consistent brand voice, and more accurate SEO and automation outputs. The Brand Vault and Knowledge Graph are the foundation of how Search Atlas understands your business. By managing this data through MCP, you ensure that every action — from content generation to SEO optimization — is aligned with your brand and business reality. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Data Retrieval & Read-Only Actions

Read-only actions allow your AI assistant to access existing data — such as reports, backlinks, usage, and analytics — without generating new work. This article explains what read-only actions are, how they work, when to use them, and how they differ from execution-based actions. 🧠 What Read-Only Actions Are Read-only actions are MCP operations that retrieve existing data, display reports or analytics, and inspect current state (projects, campaigns, profiles). They do not create, modify, or submit anything. Because of this, they do not consume Search Atlas quota — they only consume AI assistant tokens. 🔍 What You Can Retrieve Using MCP read-only actions, you can access: - Backlink data - Keyword data - SEO reports - GBP locations and data - Campaign performance - Usage and quota information - Previously generated content These actions allow you to explore and analyze your data before deciding to execute changes. ⚙️ Retrieve Data Using MCP 1. Open your MCP-connected AI assistant. 2. Ask for data or insights, for example: "Show me my backlinks," "Check my quota," "Show my GBP locations," or "View my SEO audit results." 3. The assistant retrieves existing data — no new work is created, the MCP simply returns stored or previously generated information. 4. Review the results to analyze performance, identify issues, and plan next steps. 🔁 Use Read-Only Before Execution 1. Start with a read-only request, for example: "Show me my SEO issues." 2. Review the data. 3. Decide on the next action, for example: "Apply fixes" or "Generate recommendations." 4. Execute only after review — this ensures controlled and intentional usage. ⚖️ Read-Only vs Work-Producing Actions - Read-only actions retrieve data, do not consume quota, and are used for analysis. - Work-producing actions generate, create, or submit, do consume quota, and are used for execution. 📊 Examples - Read-only — "Show me my backlinks": uses tokens, no quota consumed. - Execution — "Build backlinks for my site": uses tokens, consumes quota. - Hybrid workflow — "Show me keyword gaps" (read-only), then "Create content for those keywords" (execution). ⚠️ Important Notes 1. Read-only actions are safe and non-destructive. 2. They do not trigger approvals in most cases. 3. They are ideal for exploration and planning. 4. They still depend on your permissions and plan access. 🧠 Best Practices - Always begin with read-only queries before executing changes. - Use read-only actions to confirm data before running high-cost actions. - Combine read-only insights with execution to guide actions. Read-only actions are the safest and most efficient way to start working with MCP. By retrieving and analyzing data first, you can make better decisions, avoid unnecessary credit usage, and execute actions with confidence using your AI assistant. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Website Studio

Website Studio allows you to create projects, publish websites, and manage automation workflows connected to other Search Atlas products. Through MCP, your AI assistant can handle project creation, publishing, and synchronization tasks. This article explains how to create Website Studio projects, publish them, and manage their lifecycle using MCP. 🧠 What You Can Do Using MCP with Website Studio, you can: - Create Website Studio projects - List and manage projects - Publish websites - Monitor credit status - Sync projects with OTTO PPC - Manage automation workflows Website Studio is focused on project creation and automated website management. ⚙️ Create a Website Project 1. Open your MCP-connected AI assistant. 2. Ask to create a project, for example: "Create a Website Studio project." 3. The assistant creates the project. 4. Confirm the project details — this establishes your website project. 📋 View and Manage Projects 1. Ask your assistant, for example: "Show my Website Studio projects." 2. The assistant retrieves the project list. 3. Review the available projects. 📤 Publish a Website 1. Ask your assistant, for example: "Publish my website project." 2. The assistant prepares the publication. 3. Approve the action, if required. 4. The website is published and made live. 📊 Monitor Credit Status 1. Ask your assistant, for example: "Check my Website Studio credits." 2. The assistant retrieves credit data. 3. Review available credits and usage status. 🔄 Sync with OTTO PPC 1. Ask your assistant, for example: "Sync this project with OTTO PPC." 2. The assistant initiates synchronization. 3. Confirm the sync. 🔁 Manage Automation Workflows 1. Ask your assistant, for example: "Manage my Website Studio automation." 2. The assistant retrieves automation data. 3. Update or manage the workflows. ⚠️ Important Notes 1. Creating and publishing projects may consume credits. 2. Publishing is an external-facing action. 3. Some actions may require approval. 4. Integration with other tools depends on plan access. 🔁 How This Fits Into MCP Workflows A typical workflow: create the project, configure and manage it, publish the website, monitor credits and status, then sync with other tools. 🧠 Best Practices - Review project setup before publishing. - Monitor credit usage regularly. - Use integrations to connect workflows. - Manage automation carefully. MCP for Website Studio allows you to create and manage websites through conversation. From project creation to publishing and automation, your AI assistant can execute website workflows using the same capabilities available in Search Atlas. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Digital PR

The Digital PR system enables email-based outreach to publishers and prospects. Through MCP, your AI assistant can create campaigns, manage templates, send outreach, and handle replies. This article explains how to set up outreach campaigns, manage communication, and track responses using MCP. 🧠 What You Can Do Using MCP with Digital PR, you can: - Create outreach campaigns - Build and manage email templates - Identify outreach opportunities - Schedule and send emails - Monitor inbox and replies - Manage connected email accounts ⚙️ Create an Outreach Campaign 1. Open your MCP-connected AI assistant. 2. Ask to create a campaign, for example: "Create a digital PR campaign" or "Start an outreach campaign." 3. The assistant sets up the campaign. 4. Review the campaign setup. ✉️ Create Email Templates 1. Ask your assistant, for example: "Create an outreach email template." 2. The assistant generates the template. 3. Review and refine the content — templates are used for outreach emails. 🎯 Identify Opportunities 1. Ask your assistant, for example: "Find outreach opportunities." 2. The assistant retrieves potential prospects. 3. Review the opportunities. 📤 Send Outreach Emails 1. Ask your assistant, for example: "Send outreach emails." 2. Confirm recipients and the template. 3. The assistant schedules or sends the emails. 📬 Monitor Replies 1. Ask your assistant, for example: "Show campaign replies" or "Check my outreach inbox." 2. The assistant retrieves replies. 3. Review the responses. 🔁 Manage Campaigns 1. Ask your assistant, for example: "Manage my outreach campaign." 2. The assistant retrieves campaign data. 3. Make updates. 4. Continue or adjust the campaign. ⚠️ Important Notes 1. Outreach involves external communication. 2. Email sending may require connected accounts. 3. Campaign success depends on recipient engagement. 4. Actions may consume quota depending on execution. 🔁 How This Fits Into MCP Workflows A typical workflow: create the campaign, build templates, identify prospects, send outreach, then monitor replies. This enables full outreach campaign management through conversation. 🧠 Best Practices 1. Review templates before sending. 2. Target relevant prospects. 3. Monitor replies regularly. 4. Adjust campaigns based on responses. MCP for Digital PR allows you to run outreach campaigns through conversation. From creating templates to sending emails and managing replies, your AI assistant can execute the full campaign lifecycle using the same capabilities available in Search Atlas. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP Endpoints & Namespaces

The recommended customer-facing endpoint is https://mcp.searchatlas.com/mcp/. This root endpoint aggregates all Search Atlas namespaces and exposes the full catalog of available MCP tools. 🌐 The Main MCP Endpoint 1. Use the root endpoint: https://mcp.searchatlas.com/mcp/ 2. Add this endpoint to your MCP-compatible client. Supported clients include Claude Desktop, Claude.ai, Cursor, and ChatGPT connectors. 3. Complete the Search Atlas login flow. Once connected, your assistant can access the Search Atlas tools available to your account, based on your plan entitlements and quotas. 🧩 What Namespaces Are The MCP also exposes per-product namespaces. Examples include /mcp/otto/ and /mcp/local-seo-gbp/. A namespace scopes access to a specific product area instead of exposing the full tool catalog. ✅ Which Endpoint Should Customers Use? Customers should use the root endpoint, https://mcp.searchatlas.com/mcp/. The root endpoint aggregates all namespaces and is what customers should use unless they have a specific reason to scope access. 🔁 Root Endpoint vs Namespace Endpoint - Root endpoint (/mcp/) — exposes every tool across every product area. Recommended for customers. - Namespace endpoint, such as /mcp/otto/ — exposes only product-specific tools plus shared common tools. Recommended only when there is a specific reason to scope access. 🧠 Shared Tools Across Namespaces Some common tools are available in every namespace, including Brand Vault, Knowledge Graph, and Billing. This means even scoped namespaces may still include shared platform-level tools. ⚠️ Important Notes - The root endpoint is the standard setup path. - Namespace endpoints exist for scoped access. - Using a namespace may reduce the available tool surface. - Customers should use the root endpoint unless instructed otherwise. For most users, the correct MCP setup is simple: use https://mcp.searchatlas.com/mcp/, complete OAuth login, and let your AI assistant access the Search Atlas tools available under your account. Namespace endpoints are only needed when access should be limited to a specific product area. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Account & Billing

The Account & Billing tools give your AI assistant visibility into your Search Atlas account so it can help you track subscription status, usage, and billing without you having to log in and check manually. This article explains how to check your account status, monitor usage, and review billing details through MCP. 🧠 What You Can Do Using MCP for Account & Billing, you can: - View account information - Check subscription status - Monitor quota and usage - View billing balances - Review payment-related details 👉 These tools provide visibility into your account and usage across Search Atlas. ⚙️ Step-by-Step: View Account Information Step 1. Open your MCP-connected AI assistant. Step 2. Ask for account details, for example: "Show my account information." Step 3. The assistant retrieves your account data. Step 4. Review the account details returned. 💳 Step-by-Step: Check Subscription Status Step 1. Ask your assistant: "What is my subscription status?" Step 2. The assistant retrieves your subscription data. Step 3. Review your active plan and subscription status. 📊 Step-by-Step: Monitor Usage and Quota Step 1. Ask your assistant: "Check my usage" or "Show my quota." Step 2. The assistant retrieves your usage data. Step 3. Review your quota usage and remaining credits. 💰 Step-by-Step: View Billing Balance Step 1. Ask your assistant: "Show my billing balance." Step 2. The assistant retrieves your balance data. Step 3. Review your account balance and payment status. 🧾 Step-by-Step: Review Payment Information Step 1. Ask your assistant: "Show my billing details." Step 2. The assistant retrieves your billing information. Step 3. Review payment-related data and billing records. ⚠️ Important Notes 1. Billing data is read-only through MCP. 2. Viewing account data does not consume Search Atlas quota. 3. Payments and billing changes may require action outside of MCP. 4. Data reflects your current Search Atlas account status. 🔁 How This Fits Into MCP Workflows A typical workflow looks like this: 1. Check account status. 2. Monitor quota and usage. 3. Review billing balance. 4. Confirm readiness before executing actions. 👉 This helps prevent interruptions caused by quota or billing issues. 🧠 Best Practices 1. Check quota before running large actions. 2. Monitor billing regularly. 3. Confirm subscription status when features are unavailable. 4. Use account data to plan your usage. MCP for Account & Billing gives you visibility into your Search Atlas account through conversation. By checking usage, subscription status, and billing details, your AI assistant helps you stay in control of your resources while using the platform's capabilities. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Press Releases

The Press Release system allows you to draft AI-generated releases and distribute them through the Search Atlas network. Through MCP, your AI assistant can manage the full lifecycle from writing to publication and tracking. This article explains how to generate, publish, and manage press releases using MCP. 🧠 What You Can Do Using MCP with Press Releases, you can: - Draft press releases - Generate content using AI - Select categories and distribution networks - Publish and distribute releases - Track distribution status - Monitor quota usage Premium press release distribution tiers may consume Hyperdrive Credits (HDC). ⚙️ Create a Press Release 1. Open your MCP-connected AI assistant. 2. Ask to create a press release, for example: "Write a press release for my business." 3. The assistant generates the draft. 4. Review the content. ✍️ Edit and Finalize Content 1. Ask your assistant, for example: "Edit this press release" or "Refine the content." 2. The assistant updates the draft. 3. Review the final version and ensure accuracy before distribution. 📡 Select Distribution Settings 1. Ask your assistant, for example: "Prepare this press release for distribution." 2. Specify the category and distribution network. 3. Confirm the settings. 📤 Publish and Distribute 1. Ask your assistant, for example: "Publish this press release." 2. The assistant prepares the distribution. 3. Approve the action, if required — distribution may involve external publication and potential HDC usage. 4. The press release is published. 📊 Track Distribution Status 1. Ask your assistant, for example: "Show my press release status." 2. The assistant retrieves distribution data. 3. Review publication status, distribution progress, and campaign results. ⚠️ Important Notes 1. Drafting uses AI and may consume Search Atlas quota. 2. Distribution may consume Hyperdrive Credits (HDC). 3. Publishing interacts with external networks. 4. Approval may be required before execution. 🔁 How This Fits Into MCP Workflows A typical workflow: generate the press release, edit and finalize it, select distribution settings, publish, then track results — supporting end-to-end PR execution. 🧠 Best Practices 1. Review content before publishing. 2. Choose appropriate distribution categories. 3. Monitor distribution status after publishing. 4. Use PR alongside SEO and content strategies. MCP for Press Releases allows you to manage the full lifecycle of PR campaigns through conversation. From drafting to distribution and tracking, your AI assistant can execute each step using the same capabilities available in Search Atlas, including premium distribution options where applicable. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 Understanding MCP Errors & How to Fix Them

These errors are not random — they follow predictable patterns based on quota, authentication, permissions, rate limits, or external system behavior. This article explains the most common MCP errors, what they mean, and exactly how to fix them step by step. 🧠 How MCP Errors Work Every MCP action goes through multiple checks before execution: authentication, permissions, quota, rate limits, and tool execution. 👉 If any of these checks fail, the MCP returns an error instead of executing the action. ⚙️ Step-by-Step: How to Diagnose Any MCP Error Step 1. Read the error message carefully. Step 2. Identify the error type: quota error, authentication error, permission error, rate limit error, or execution error. Step 3. Follow the fix for that specific error category. 🔍 Common MCP Errors & Fixes ❌ Error Type 1 — Quota Exhausted What happens: The action does not run, and the assistant reports insufficient credits. Why this happens: Your Search Atlas quota is fully used. How to fix: 1. Ask: "Check my quota." 2. Confirm usage is at limit. 3. Add credits or upgrade your plan. 4. Retry the action. ❌ Error Type 2 — Invalid or Expired Token What happens: You see an "Invalid token" error and authentication fails. Why this happens: Your session is expired or near expiration. How to fix: 1. Disconnect MCP in your client. 2. Reconnect MCP. 3. Complete login again. ❌ Error Type 3 — Permission / Access Denied What happens: The action fails immediately and the tool cannot execute. Why this happens: Your plan does not include the requested feature. How to fix: 1. Check your Search Atlas plan. 2. Confirm feature access. 3. Upgrade or enable the feature. 4. Retry the action. ❌ Error Type 4 — Rate Limit (429 Error) What happens: The request is rejected and the error message includes "429." Why this happens: Too many requests were sent in a short time. How to fix: 1. Stop sending requests. 2. Wait briefly. 3. Retry the request. 4. Reduce request frequency. ❌ Error Type 5 — Action Requires Approval What happens: The action does not run immediately and the system waits. Why this happens: The action is marked as requiring approval before it can run. How to fix: 1. Look for the approval prompt. 2. Review the action. 3. Approve it to proceed. ❌ Error Type 6 — External System Failure What happens: An integration action fails, with partial execution or no result. Why this happens: The external system isn't connected, an API failure occurred, or credentials are invalid. How to fix: 1. Check the integration connection. 2. Verify account access. 3. Retry the action. ❌ Error Type 7 — Tool Execution Failure What happens: The action starts but fails mid-execution. Why this happens: Invalid input, an incomplete request, or an upstream processing issue. How to fix: 1. Review your request. 2. Make it more specific. 3. Retry with clearer instructions. 🔁 Universal Debugging Workflow If you are unsure what failed, follow this exact sequence: 1. Check quota. 2. Check authentication. 3. Check permissions. 4. Check for an approval prompt. 5. Retry the action. ⚠️ Important Notes 1. MCP errors are predictable and structured. 2. Most issues fall into 4 categories: quota, authentication, permissions, and rate limits. 3. Approval and integration issues are secondary causes. 🧠 Best Practices to Avoid Errors Check before executing: Ask "Will this consume credits?" Use read-only first: Analyze before executing. Keep sessions active: Avoid token expiration. Be specific in requests: Clear input reduces failures. MCP errors are not random — they are signals that one part of the execution pipeline needs attention. By identifying the error type and following the correct fix, you can quickly resolve issues and continue working efficiently through your AI assistant, using the same Search Atlas capabilities, permissions, and quotas tied to your account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP Supported Clients & Environments

This means you are not limited to a single interface — you can connect MCP across multiple environments depending on how you prefer to work. This article explains which clients are supported, how they differ, and how MCP behaves across environments. 🧠 What "Supported Client" Means A supported client is any application that can connect to an MCP endpoint, authenticate with Search Atlas, and execute MCP tool calls. 👉 MCP is not tied to a single UI — it is a protocol-based system. 🔍 Supported Clients Search Atlas currently supports the following MCP-compatible clients: - Claude Desktop - Claude.ai - Cursor - ChatGPT connectors All of these can connect to the MCP endpoint: https://mcp.searchatlas.com/mcp/ ⚙️ Step-by-Step: Connect in Any Client Step 1. Open your MCP-compatible client. Step 2. Add a new MCP server. Step 3. Use the endpoint https://mcp.searchatlas.com/mcp/. Step 4. Start the connection. Step 5. Complete the OAuth login in your browser. Step 6. Return to the client. Your MCP connection is now active. 🖥️ How Different Clients Are Used Claude Desktop / Claude.ai — Best for conversational workflows, non-technical users, and general usage. Cursor — Best for developers working inside code editors and integrating MCP into development workflows. ChatGPT Connectors — Best for using MCP within ChatGPT-supported environments. 🔁 Same System, Different Interface Regardless of the client, the same MCP endpoint is used, the same authentication applies, the same tools are available, and the same quotas and permissions are enforced. 👉 Only the interface changes — not the capabilities. ⚠️ Important Notes 1. MCP behavior is consistent across all clients. 2. Differences are only in how actions are triggered. 3. Approval flows depend on client support (for example, Claude supports human-in-the-loop prompts). 🛠️ Troubleshooting Across Clients Issue — Works in one client but not another. Possible causes: not authenticated in that client, MCP not connected, or session expired. Step-by-step fix: 1. Reconnect MCP in the affected client. 2. Complete OAuth login again. 3. Retry the action. 🔁 Best Practice 💡 Choose your client based on workflow: use Claude for ease of use, or Cursor for development workflows. All clients behave the same once connected. MCP is designed to be flexible across environments. Whether you are working in a conversational UI or a code editor, you can access the same Search Atlas capabilities through your AI assistant, using the same permissions, quotas, and system behavior tied to your account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Indexer

The Indexer allows you to send batches of URLs for indexing, check submission status, and retrieve reports. Through MCP, your AI assistant can manage the full submission workflow. This article explains how to submit URLs, monitor indexing status, and manage indexing reports using MCP. 🧠 What You Can Do Using MCP with the Indexer, you can: - Submit batches of URLs for indexing - Check submission status - View submission details - Download indexing reports 👉 Indexing actions are processed through Search Atlas's indexing service, which handles submission to search engines on your behalf. ⚙️ Step-by-Step: Submit URLs for Indexing Step 1. Open your MCP-connected AI assistant. Step 2. Ask to submit URLs, for example: "Submit these URLs for indexing." Step 3. Provide the list of URLs. Step 4. The assistant submits the batch. 👉 This action requires your confirmation before it runs, since it uses indexing quota. 🔄 Step-by-Step: Check Submission Status Step 1. Ask your assistant: "Check indexing status." Step 2. The assistant retrieves the batch status. Step 3. Review submission progress and indexing updates. 📊 Step-by-Step: View Submission Details Step 1. Ask your assistant: "Show indexing details for my URLs." Step 2. The assistant retrieves detailed data. Step 3. Review individual URL status and submission results. 📥 Step-by-Step: Download Reports Step 1. Ask your assistant: "Download indexing report." Step 2. The assistant prepares the report. Step 3. Review the report data. 👉 Reports summarize indexing activity and results. ⚠️ Important Notes 1. Submitting URLs consumes Indexer quota. 2. Viewing status and reports is read-only. 3. Indexing results depend on the receiving search engine's own processing. 4. Submissions interact with external systems. 🔁 How This Fits Into MCP Workflows A typical workflow: 1. Submit URLs. 2. Monitor indexing status. 3. Review reports. 4. Repeat as needed. 👉 This supports ongoing indexing management. 🧠 Best Practices 1. Submit important or updated URLs. 2. Monitor indexing progress regularly. 3. Use reports to track results. 4. Combine with SEO workflows for best outcomes. MCP for Indexer allows you to manage URL submissions through conversation. By submitting batches, tracking status, and reviewing reports, your AI assistant can handle indexing workflows using the same capabilities available in Search Atlas. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔗 Access Backlink Data via Search Atlas MCP/API

🔍 Overview Many customers building custom client dashboards want to populate sections like Top Referring Domains and Recent Backlinks using the Search Atlas MCP or API. This article explains the current state of backlink data access, what tools are available, and how to work effectively with the data while improvements are in progress. 📊 What Backlink Data Is Currently Available Search Atlas provides backlink and referring domain data through the platform's Authority section. You can access this data manually by navigating to Left sidebar → Authority → Campaigns to review link-building activity, or explore related tools under the Authority menu. The following MCP tools exist for programmatic backlink access: - se_get_backlinks — intended to return backlink-level data for a target domain - se_get_referring_domains — intended to return a list of referring domains pointing to a target domain ⚠️ Known Issue: MCP Backlink Tools Returning No Data There is a known active issue where se_get_backlinks and se_get_referring_domains return no data on all tested domains. Our engineering team is actively investigating this bug. Until the fix is deployed, these two MCP tools are not reliable for populating custom dashboards. This is not a configuration error on your end. If you are calling these tools and receiving empty responses, the issue is on the platform side and is being resolved. 🛠️ Workarounds While the Fix Is in Progress While the MCP backlink tools are under repair, consider the following options to keep your custom dashboards functional: 1. Export data manually from the platform. Navigate to Left sidebar → Authority → Campaigns to review backlink campaign data. Use any available export options to pull data into your reporting workflow. 2. Use the URL Indexer for link visibility. Go to Left sidebar → Authority → Indexer (URL Indexer) to check indexation status of backlink URLs you already know about. 3. Combine with other available API endpoints. If your dashboard includes additional SEO metrics such as keyword rankings or organic traffic, those endpoints may already be functional. Build around the working data points while backlink tools are restored. 4. Monitor the platform for updates. Once the engineering fix is deployed, se_get_backlinks and se_get_referring_domains will return data as expected and you can integrate them directly into your dashboard pipeline. 🚀 What to Expect After the Fix Once the issue is resolved, you will be able to use the MCP tools as follows: - Call se_get_referring_domains with a target domain to retrieve a list of domains linking to it — ideal for populating a Top Referring Domains table in your dashboard. - Call se_get_backlinks with a target domain or URL to retrieve individual backlink records — ideal for a Recent Backlinks feed or table. Both tools are designed to return structured data that can be mapped directly to dashboard components. No additional configuration should be required once the data pipeline is restored. 💡 Tips for Building a Reliable Custom Dashboard - Build your dashboard with graceful empty states for backlink sections so incomplete data does not break the layout for clients. - Add error handling around MCP tool calls so your integration can distinguish between an empty result and a failed request. - Use caching where possible — backlink data does not change minute to minute, so storing the last successful response protects your dashboard during any future API interruptions. 🙋 Need Further Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔌 What Is MCP and Where to Find It

🧩 What Is MCP? MCP stands for Model Context Protocol. It is an open standard that allows AI assistants and external tools to connect directly to platforms like Search Atlas and perform actions on your behalf — such as retrieving data, managing settings, or triggering account-level operations — through a secure, structured connection. Think of MCP as a bridge between an AI agent (for example, Claude, Cursor, or another MCP-compatible client) and your Search Atlas account. Instead of copying and pasting data manually, an AI agent uses an MCP connection to read and interact with Search Atlas in real time. 🔑 How MCP Differs from a Standard API Key Both MCP connections and standard API keys allow external tools to access your Search Atlas data, but they work differently: - Standard API key: A static credential you embed in code or third-party integrations. It authenticates requests but does not carry session context or support interactive, multi-step workflows. - MCP connection: A dynamic, session-aware protocol designed specifically for AI agents. It supports back-and-forth interactions, tool calls, and structured responses that an AI model can act on directly. MCP connections are better suited for automated workflows that require multiple sequential actions. Use a standard API key when integrating Search Atlas data into a custom script or third-party app. Use an MCP connection when you want an AI assistant to operate Search Atlas tools on your behalf interactively. 🗺️ Where to Find MCP Connections in Search Atlas MCP connection setup and authentication are managed from within your Search Atlas account settings area. Because the exact location of MCP connection options may vary depending on your plan and the current dashboard layout, please contact our support team via the chat widget so they can confirm the correct navigation path for your account and verify that MCP access is included in your current subscription. To get the fastest resolution, please have the following ready when you reach out: - Your Search Atlas account email or project name - Your current subscription plan - A description of what you are trying to connect (e.g., which AI agent or tool you are using) - Any error messages or authentication failures you have encountered, including the exact wording and any timestamps ⚙️ When and Why to Use MCP MCP connections are most useful in the following situations: - AI-assisted SEO workflows: Connect an AI agent to Search Atlas so it can pull keyword data, check rankings, or analyse site metrics without manual exports. - Automated reporting: Let an AI agent query Search Atlas tools and compile results into a report automatically. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP API Documentation

The MCP API Documentation provides full technical details for connecting and interacting with Search Atlas via the Model Context Protocol, including authentication, endpoints, and integration setup. 🔗 Click Here To Access MCP API Documentation ⚠️ Important - This documentation is intended for technical users. - All MCP configuration, capabilities, and implementation details are covered in the official docs. - Support for MCP or API implementation is not provided through this knowledge base. ✅ Next Step To proceed, open the MCP API documentation using the link above and follow the integration instructions provided. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🤖 MCP Tools vs Slash Commands in Claude Code

🔍 What This Article Covers If you attended the Search Atlas AI summit, you may have seen a demo featuring 60+ AI-powered tools. When you open Claude Code, however, you only see around 20 items listed — and /onboard-client is nowhere to be found. This article explains exactly why, and how to access every tool available to you. ⚡ The Two Types of Tools: A Critical Distinction There are two completely separate categories of tools in the Search Atlas AI ecosystem. Confusing them is the most common source of frustration, so understanding this distinction will clear up everything. - Slash commands — These are built-in Claude Code shortcuts that you type manually. You trigger them yourself by typing a forward slash. Examples include /help, /clear, and /cost. Claude Code ships with roughly 20 of these commands, and that number is expected and correct. - MCP tools — These are the 60+ Search Atlas-powered capabilities (keyword research, site audits, content creation, SERP analysis, and more). You do not type a slash command to use them. Instead, you simply describe what you want in plain language during your conversation and the AI model automatically selects and invokes the right MCP tool on your behalf. MCP tools are never listed as slash commands. They run silently in the background when the model determines they are needed. No manual loading, syncing, or selection is required. 💬 How to Use MCP Tools in Claude Code Because MCP tools are invoked automatically through natural language, using them is straightforward. You do not need to remember command names or a special syntax. 1. Open Claude Code and confirm that your Search Atlas MCP server connection is active. 2. Start a new conversation and describe your task in plain English. For example: "Run a technical site audit for example.com" or "Find keyword opportunities for a blog post about home insurance." 3. Claude Code will automatically detect which MCP tool is appropriate and execute it. You will see a tool-call notification in the conversation thread when a tool is being used. 4. Review the results returned directly in the conversation. You can ask follow-up questions or request additional actions in the same thread. There is nothing else to install, enable, or unlock. If your MCP server connection is active, all 60+ tools are already available and ready to be called by the model. 📋 Why You See ~20 Items in the Slash Command List When you type / in Claude Code, the autocomplete panel shows only Claude Code's native built-in commands. This list will always contain roughly 20 items regardless of how many MCP tools are connected. This is intentional behavior — MCP tools are a completely different integration layer and are not surfaced in that panel. Seeing ~20 slash commands does not mean your MCP tools are missing, broken, or incomplete. 🌐 Where to Find /onboard-client /onboard-client is not part of the MCP toolset and is not available inside Claude Code. It is a dedicated feature of the Search Atlas AI Agent web interface, which is a separate product accessed at https://agent.searchatlas.com. To use /onboard-client and similar workflow commands: 1. Navigate to https://agent.searchatlas.com in your browser. 2. Log in with your Search Atlas credentials. 3. Type /onboard-client directly in the AI Agent chat interface. The AI Agent web interface and Claude Code are complementary tools with overlapping but distinct feature sets. Features specific to the web interface will not appear in Claude Code, and vice versa. ✅ Quick Reference Summary - ~20 slash commands in Claude Code — Normal and expected. These are Claude Code's built-in utilities. - 60+ MCP tools — Available in Claude Code via natural language conversation. Not listed as slash commands. - /onboard-client — Only available at https://agent.searchatlas.com. Not part of Claude Code or the MCP toolset. - To trigger an MCP tool — Just describe your goal in plain language. The model handles the rest automatically. 🛠️ Troubleshooting: MCP Tools Not Responding If you describe a task and the model does not appear to be using any MCP tools, check the following before escalating: - Confirm your Search Atlas MCP server is listed as connected in your Claude Code MCP settings panel. - Make sure you are using a Claude model that supports tool use (Claude 3.5 Sonnet or later is recommended). - Try rephrasing your request to be more specific about the outcome you want. Vague prompts are less likely to trigger a tool call. - Check that your Search Atlas account has an active subscription with API access enabled. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Google Business Profile (Local SEO)

Use MCP to manage your Google Business Profile (GBP) locations, content, and optimization workflows through the Local SEO product area. The MCP provides full Google Business Profile management, including locations, services, hours, attributes, categories, posts, reviews, media uploads, and audit reports. 🧠 What You Can Do Using MCP for Local SEO (GBP), you can: - Manage locations - Update categories, attributes, and services - Create and publish posts - Respond to reviews with AI-generated replies - Upload and manage media - Run audit reports - Submit and track citations ⚙️ Step-by-Step: Manage GBP Locations 📌 Step 1 Open your MCP-connected AI assistant 📌 Step 2 Ask to view your locations Examples: - “Show me my GBP locations” - “List all my business locations” 📌 Step 3 Review your locations The assistant returns your available locations and their details. 🧩 Step-by-Step: Update Categories, Attributes & Services 📌 Step 1 Ask your assistant: - “Update my GBP categories” - “Add attributes to my location” - “Update services for my business” 📌 Step 2 The assistant prepares the updates 📌 Step 3 Confirm changes (if required) 📌 Step 4 The MCP applies updates to your GBP profile 📝 Step-by-Step: Create & Publish Posts 📌 Step 1 Ask your assistant: - “Create a GBP post for my business” - “Publish a promotion post” 📌 Step 2 The assistant generates the post content 📌 Step 3 Review the draft 📌 Step 4 Approve and publish the post Posts are then pushed to your Google Business Profile. 💬 Step-by-Step: Respond to Reviews 📌 Step 1 Ask your assistant: - “Show me my GBP reviews” - “Respond to my latest review” 📌 Step 2 The assistant retrieves reviews 📌 Step 3 The assistant generates a reply 📌 Step 4 Approve the response The reply is then published to your GBP listing. 📸 Step-by-Step: Manage Media 📌 Step 1 Ask your assistant: - “Upload images to my GBP profile” - “Show my GBP media” 📌 Step 2 The assistant processes the request 📌 Step 3 Media is uploaded and associated with your listing 📊 Step-by-Step: Run GBP Audit Reports 📌 Step 1 Ask your assistant: - “Run a GBP audit for my location” 📌 Step 2 The assistant generates the audit 📌 Step 3 Review results and recommendations 🌐 Step-by-Step: Manage Citations 📌 Step 1 Ask your assistant: - “Create citations for my business” - “Submit my business to directories” 📌 Step 2 The assistant prepares citation data 📌 Step 3 Submit to selected networks 📌 Step 4 Track progress and status ⚠️ Important Notes - Actions like audits, content generation, and submissions consume Search Atlas quota - Publishing and updating existing content does not consume additional quota - Some actions may require confirmation before execution 🔁 How This Fits Together The GBP workflow through MCP typically follows: 1. Manage locations 2. Optimize categories, attributes, and services 3. Create posts and content 4. Respond to reviews 5. Upload media 6. Run audits and manage citations All actions are executed through your AI assistant using the same system as the Search Atlas platform. MCP enables full Google Business Profile management through conversation. From optimizing your listing to publishing content and managing reviews, your AI assistant can execute every step using the same tools, permissions, and quotas available in your Search Atlas account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for Report Builder

Report Builder allows you to generate custom reports using data from multiple Search Atlas product areas. Through MCP, your AI assistant can create reports, manage templates, and retrieve report outputs. This article explains how to create reports, manage templates, and retrieve reporting data using MCP. 🧠 What You Can Do Using MCP with Report Builder, you can: - Create custom reports - Use and manage report templates - Pull data from multiple product areas - Generate and retrieve reports - Manage reporting workflows Reports consolidate data across tools into a single output. ⚙️ Create a Report 1. Open your MCP-connected AI assistant. 2. Ask to create a report, for example: "Create a report" or "Generate a custom SEO report." 3. The assistant prepares the report. 4. Confirm the report setup — reports can include data from multiple product areas. 🧩 Use Report Templates 1. Ask your assistant, for example: "Show available report templates." 2. The assistant retrieves the templates. 3. Select a template. 4. Apply the template to generate the report. 📊 Generate Report Data 1. Ask your assistant, for example: "Generate report data." 2. The assistant compiles data. 3. Review the report output — data is aggregated from multiple sources. 📥 Retrieve Reports 1. Ask your assistant, for example: "Show my reports." 2. The assistant retrieves the report list. 3. Select a report. 4. Review the report details. 🔁 Manage Reports 1. Ask your assistant, for example: "Manage my reports." 2. The assistant retrieves report data. 3. Update or regenerate reports as needed. ⚠️ Important Notes - Generating reports may consume Search Atlas quota. - Viewing existing reports is read-only. - Report content depends on available data sources. - Reports combine data across multiple tools. 🔁 How This Fits Into MCP Workflows A typical workflow: select or create a report, apply a template, generate report data, review the output, then update as needed — supporting consolidated performance reporting. 🧠 Best Practices 1. Use templates for consistency. 2. Verify included data sources. 3. Regenerate reports after major updates. 4. Use reports to guide decisions. MCP for Report Builder allows you to generate and manage reports through conversation. By combining data from multiple Search Atlas tools into structured outputs, your AI assistant can help you monitor performance and track results using the same capabilities available in the platform. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🛠️ MCP Tools vs Slash Commands in Claude Code: Why You See Fewer Commands

Overview If you are using MCP (Model Context Protocol) access with Claude Code and notice that the number of visible slash commands is lower than the total number of capabilities you expected, you are not missing features — you are looking at two different systems. This article explains the difference between MCP tool invocation and slash commands in Claude Code. MCP Tools vs Slash Commands: The Key Difference Many customers assume that every capability should appear as a slash command. In reality, Claude Code exposes functionality through two distinct mechanisms: - Slash commands are shortcuts you type manually in the Claude Code client. These are a curated set of commands surfaced directly in the interface. - MCP tools are invoked automatically by the agent when it determines a task requires them. They do not appear in your slash command list, but they are still active and available in the background. This is why the number of visible slash commands may appear much lower than the total number of MCP tools available. The MCP tools operate in the background and are not surfaced as typed commands in the menu. Why You May See Fewer Commands There are a few common reasons the visible command count may differ from what you expected: - MCP tools are auto-invoked, not listed: A significant portion of available capabilities are MCP tools that the agent calls automatically based on the task at hand. These will not appear in the slash command menu. - Slash commands are a separate, curated set: The slash commands visible in your client represent a distinct subset of functionality, not the full list of everything the agent can do via MCP tools. Still Have Questions? If you are unsure whether your MCP setup is working as expected or have additional questions about how tools and commands are configured for your account, use the chat icon in the bottom-right corner of the platform and type human teammate to be connected with a member of our team and we will be happy to help. Additional Notes Candidate specifies ~20 slash commands is expected/correct, names examples (/help, /clear, /cost), mentions 60+ MCP tools, and gives step-by-step usage: describe task in plain English, e.g. 'Run a technical site audit for example.com'; Claude auto-selects the tool; tool-call notification appears in thread.

🔄 WordPress User Refresh and MCP Data Access

🗺️ Overview This article covers two common questions about Search Atlas integrations: how to refresh the list of WordPress users available for content publishing, and what analytics data can be accessed through the Model Context Protocol (MCP) for Claude. Both topics are handled separately below. 👤 Refreshing WordPress Users for Content Publishing When you connect a WordPress site to Search Atlas, the platform pulls in a list of authors and users from that site. If you add new WordPress users after the initial connection, you will need to manually refresh the list so Search Atlas recognises them for content publishing assignments. ✅ Prerequisites - An active WordPress site connected to Search Atlas via the WordPress integration. - The Search Atlas plugin installed and activated on your WordPress site. - Administrator-level access on both your WordPress site and your Search Atlas account. - Any new users you want to appear must already exist in WordPress before you refresh. 🔧 How to Refresh WordPress Users The exact location of the WordPress user refresh option may vary depending on your account configuration. To find and use it: 1. Log in to your Search Atlas account. 2. Go to your integrations or settings area and open your WordPress integration. 3. Look for a refresh or re-sync option associated with your connected WordPress site and select it. 4. Search Atlas will re-fetch the user list from your WordPress site. This typically completes within a few seconds. 5. Once the refresh is complete, your updated user list will be available when assigning authors to content tasks and publishing workflows. Note: If a user does not appear after refreshing, confirm that the user account exists and is active in WordPress, and that the Search Atlas plugin has the correct API credentials to communicate with your site. If you are unsure where to find the refresh option in your account, please reach out to our support team for step-by-step guidance. 📊 GA4 and Google Ads Data Through MCP for Claude The Search Atlas MCP (Model Context Protocol) integration allows Claude to access analytics and advertising data from your connected accounts. Because the specific data fields and reports available through MCP can vary by account configuration and may be updated over time, we recommend contacting our support team to confirm exactly which GA4 and Google Ads data points are accessible for your setup and how to use them in custom client reporting workflows. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 MCP for SEO & Site Audits (OTTO)

Use MCP to run SEO audits, generate recommendations, and deploy optimizations using the OTTO product area. OTTO is the Search Atlas system for site audits, SEO recommendations, schema management, and indexing. Through MCP, your AI assistant can execute the full audit and optimization workflow on your behalf. 🧠 What You Can Do Using MCP with OTTO, you can: - Run full site audits - Generate prioritized SEO recommendations - Review issues and summaries - Edit and deploy fixes - Manage schema markup - Submit URLs for indexing ⚙️ Step-by-Step: Run an SEO Audit 📌 Step 1 Open your MCP-connected AI assistant 📌 Step 2 Ask for an audit Examples: - “Run a site audit for my website” - “Audit my domain and show me SEO issues” 📌 Step 3 The assistant creates an audit and retrieves results The MCP runs a full site audit and returns issue summaries and recommendations. 📌 Step 4 Review the results You will see: - SEO issues - priority levels - recommended fixes 🛠️ Step-by-Step: Generate SEO Recommendations 📌 Step 1 Ask your assistant: - “Generate SEO recommendations for my site” - “Show me fixes for my audit issues” 📌 Step 2 The assistant generates recommendations These are prioritized and based on your audit data. 📌 Step 3 Review the recommendations Each recommendation includes suggested actions to improve SEO performance. 🚀 Step-by-Step: Deploy Fixes 📌 Step 1 Ask your assistant: - “Apply SEO fixes to my site” - “Deploy the recommended changes” 📌 Step 2 The assistant prepares the deployment 📌 Step 3 Confirm the action (if required) Some actions may require approval before execution. 📌 Step 4 The MCP deploys the fixes Changes are applied using the same system as the Search Atlas app. 🧱 Step-by-Step: Manage Schema & Indexing 🧱 Schema Management Ask: - “Add schema markup to my pages” - “Update my site schema” The assistant can manage page-level and domain-level schema. 📥 Indexing Ask: - “Submit these URLs for indexing” - “Index my new pages” The MCP submits URLs through the indexing system. ⚠️ Important Notes - Audits, recommendations, and indexing actions consume Search Atlas quota - Deploying already-generated fixes does not consume additional quota - Some actions may require confirmation before execution 🔁 How This Fits Together The OTTO workflow through MCP follows this sequence: 1. Run audit 2. Generate recommendations 3. Review issues 4. Deploy fixes 5. Manage schema and indexing All steps are executed through your AI assistant using the same system as the Search Atlas platform. MCP allows you to manage your entire SEO audit and optimization workflow through conversation. From identifying issues to deploying fixes and indexing pages, your AI assistant can execute each step using the same tools, permissions, and quotas available in your Search Atlas account. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🚧 Ad Studio MCP Agency Workflow Blockers Explained

Overview If you are using the Ad Studio MCP (Model Context Protocol) to automate agency workflows, you may encounter limitations that affect multi-account analytics and creative deployment. This article explains what is happening and what to have ready when contacting our team. What Is Happening Customers using the Ad Studio MCP for agency workflows involving multiple ad accounts have reported gaps in analytics data and creative deployment functionality. What This Affects - Multi-account analytics queries run through the MCP - Creative deployment tasks automated via MCP What You Should Do Now Because these are platform-level limitations rather than configuration issues, there are no self-serve steps that will fully resolve the gaps at this time. To get help as quickly as possible, please have the following ready when you contact our team: - The name of the ad account or accounts where you are experiencing the issue - A description of the specific MCP action or query that is producing unexpected results (e.g., analytics not returning, creative not deploying) - Any exact error messages or unexpected responses you have received - The approximate date and time when you first noticed the problem Having this information ready will allow our team to investigate your specific case and escalate it appropriately. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🛠️ MCP v2 API Key Requests Hanging Indefinitely

🔍 What This Article Covers Some customers using the MCP (Model Context Protocol) endpoint with a valid v2 X-API-KEY header experienced requests that hung indefinitely — never returning a response, not even an error. At the same time, requests sent without an API key returned a fast 401 Unauthorized error, which confirmed the endpoint itself was reachable. This article explains what caused the issue, confirms it has been resolved, and tells you what to do if you still experience hanging requests. ⚙️ Root Cause: Redis Timeout During v1→v2 Migration This was a confirmed systemic issue, tracked internally as SPE-1065. During the migration from the legacy v1 API key system to the new v2 system, the allow-list check — the step that validates whether a key is permitted to access the MCP endpoint — relied on a shared Redis instance. A timeout in that Redis connection caused the allow-list check to stall on every MCP pod simultaneously. The result: any request carrying a v2 X-API-KEY header would pause at the authentication step and never complete. Requests with no key bypassed that check entirely and received an immediate 401, which is why those returned quickly. This was not caused by an invalid or revoked key. The key itself was valid — the authentication infrastructure was unable to verify it due to the Redis timeout. ✅ Current Status: Resolved The engineering team identified the root cause, reviewed the fix, and deployed it to production. MCP v2 API key authentication is now functioning correctly. The allow-list check is reachable from all MCP pods and no longer times out. 🚀 Steps to Take If You Are Still Experiencing Hanging Requests 1. Confirm you are using a v2 API key. Legacy v1 keys are no longer supported for MCP endpoint access. If you are unsure which version your key is, check your API key settings in your account. 2. Retry your request. Now that the fix has been deployed, a fresh request using a valid v2 X-API-KEY header should authenticate and respond correctly. 3. Check for any network-level timeouts on your side. If your client or infrastructure has a short timeout configured, ensure it is set high enough to allow the server to respond (a few seconds is typically sufficient). 4. Verify the key has not been revoked. If your key was deactivated or flagged, requests may fail. If you suspect your key status has changed, review it in your account settings or generate a new one. 5. Test with and without the header. If requests without the key still return a fast 401 but requests with the key still hang, this is a strong signal that something is still wrong server-side — not on your end. In this case, contact our team immediately (see below). 💡 What a Hanging Request Looks Like vs. a Normal Failure - Hanging request (the bug): Request with a valid v2 X-API-KEY header — no response, no error, connection stays open indefinitely. - Normal auth failure: Request without a key — fast 401 response, connection closes immediately. - Expected behavior (post-fix): Request with a valid v2 X-API-KEY header — authenticated successfully, response received within a normal timeframe. 📋 Information to Have Ready If You Need Support If you still experience hanging requests after the fix, gather the following before contacting our team — it will significantly speed up the investigation: - A description of the exact request being made (endpoint path, headers used — redact the key value itself) - Timestamps of when the hanging requests occurred - Whether requests without the key return a fast 401 at the same time - Any error messages or timeout messages from your client or logs - A screen recording or HAR file if possible 💬 Still Need Help? If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🛠️ Fix MCP WebSocket 403 Permission Errors

🔍 What Is This Error? When using the Search Atlas MCP (Model Context Protocol) server with an AI client like Claude to manage or edit your website, you may encounter an HTTP 403 error with a message such as "You do not have permission to perform this action." This error can appear when running playbooks, accessing SEO tools, or executing any MCP-connected operation. This is an authentication and permissions issue — it means the MCP server cannot verify that your session has the correct access rights to carry out the requested action. ⚠️ Common Triggers - Your MCP API credentials are missing, expired, or were not saved correctly. - The MCP connection was set up under a different workspace or user account than the one currently active. - Your Search Atlas session timed out, invalidating the active WebSocket connection. - A recent platform update reset or changed permission scopes for MCP integrations. - The connector was configured but has not completed its authorisation handshake. 🔧 Step-by-Step Troubleshooting 1. Verify you are in the correct workspace. In the left sidebar, click Coworker. At the top of the page, confirm the workspace name shown matches the project you are working on. If not, use the workspace switcher to select the right one. 2. Refresh your setup status. Go to Coworker → Overview tab and click Refresh setup status. Wait for the status indicators to update. A green status on all items confirms your agent connection is healthy. 3. Check your MCP connector. Navigate to Coworker → Connectors tab and click Manage connectors. Locate your MCP connector. If it shows an error state or a warning icon, click into it and re-authenticate by following the on-screen prompts to re-enter your API key or OAuth credentials. 4. Re-enter your Search Atlas API key in your MCP client. Open your MCP client configuration (for example, Claude's settings or your claude_desktop_config.json file). Confirm the API key entered matches the one found in your Search Atlas account. Copy a fresh key from your account settings if needed, paste it into the config, and save. 5. Restart the MCP server connection. After updating credentials, fully quit and relaunch your AI client (e.g., Claude Desktop) so it establishes a fresh WebSocket session with the updated credentials. Stale sessions will continue to produce 403 errors even after a key update. 6. Re-run the failing playbook or tool. Return to your AI client and retry the action that produced the 403 error. If the credentials and connector are correctly configured, the request should now succeed. 💡 Tips to Prevent This Error - Always complete the full connector setup flow in Coworker → Connectors tab before using MCP tools in your AI client. - If you belong to multiple workspaces, double-check the active workspace before running any agent playbook — permissions are workspace-specific. - Rotate API keys carefully: update them in your MCP client configuration immediately after generating a new key in Search Atlas to avoid session mismatches. - Keep your AI client (Claude Desktop or similar) updated to the latest version to ensure WebSocket compatibility with the MCP server. 🚀 Still Seeing the 403 Error? If you have followed all the steps above and the error persists, it is possible that a permission scope for your account or agent has not been correctly provisioned on the backend. This requires a manual review by our team. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🔌 Connect Search Atlas MCP to Claude via OAuth

🧭 Overview The Search Atlas MCP (Model Context Protocol) connector lets you use Search Atlas tools — such as keyword research, site audits, and OTTO insights — directly inside Claude. This guide explains what OAuth authorization is, what to do if your tools are not appearing after connection, and how to get further help from our team. 🔑 What Is OAuth Authorization for MCP? OAuth is the secure login method that links your Search Atlas account to Claude. Without completing this step, Claude cannot access your Search Atlas tools, even if the MCP server appears to be connected. You must authorize the connection through your Search Atlas account before any tools become available. ⚙️ How to Connect Search Atlas MCP to Claude The exact steps to add and authorize the Search Atlas MCP connector depend on the version of Claude you are using. Follow the instructions provided in Claude's own documentation or setup wizard for adding an MCP server. When prompted, log in with your Search Atlas credentials and approve the permissions requested to grant Claude access to your Search Atlas tools. If you are unsure where to begin, our support team can walk you through the process for your specific Claude setup. 🛠️ Troubleshooting: Search Atlas Tools Not Showing Up Problem: Authorization was not completed - If the OAuth flow was interrupted or you closed the browser before finishing, disconnect the Search Atlas MCP connector and restart the authorization process from the beginning. Ensure you complete all steps without closing the browser tab during any redirects. Problem: You are on a free or limited plan - MCP tool access requires an active Search Atlas subscription. If your account is on a free trial or a limited allowance plan, authenticated tool requests may be blocked. Check your plan status inside your Search Atlas account and upgrade if needed, then retry the connection. Problem: Tools still do not appear after authorization - Disconnect the connector entirely, log out of both Claude and Search Atlas, then log back in and repeat the authorization flow. If the issue persists, contact our support team with the details listed below. 📋 What to Have Ready When Escalating If you have tried the steps above and tools are still not appearing, please have the following information ready before contacting support: - Your Search Atlas account email and the plan you are subscribed to - The exact error message or behavior you are seeing (e.g., tools not listed, authorization page not loading, connector showing as disconnected) - The version or interface of Claude you are using (e.g., Claude.ai, API, desktop app) - A timestamp or approximate time when the issue first occurred If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

🛠️ Install the Search Atlas MCP Server

🔌 What is the Search Atlas MCP server? The Search Atlas MCP server lets supported AI tools, such as Claude, connect to Search Atlas capabilities through the Model Context Protocol (MCP). 🔑 Before you begin - Make sure Claude is installed and available from your terminal. - Have an active Search Atlas API key ready. - Replace YOUR_SEARCHATLAS_API_KEY_HERE with your actual API key before running the command. ⚙️ Install MCP in Claude Open your terminal and run the following command: claude mcp add --transport http searchatlas https://mcp.searchatlas.com/mcp/ --header "X-API-KEY: YOUR_SEARCHATLAS_API_KEY_HERE" Use standard quotation marks exactly as shown. Do not include the placeholder text, brackets, or curly quotation marks around your API key. 🔄 Reconnect after installation 1. Close and reopen Claude, or reconnect the MCP server from your Claude settings. 2. Start a new conversation. 3. Ask Claude to use the Search Atlas MCP connection. 🧪 Fix authentication errors If Claude rejects your credentials, run the installation command again with the correct API key, then reconnect Claude. Authentication commonly fails when the API key is expired, copied with extra spaces, or entered with curly quotation marks. - Confirm that the key is active in Search Atlas. - Copy the key again without leading or trailing spaces. - Use the exact server URL: https://mcp.searchatlas.com/mcp/. - Make sure the header name is exactly X-API-KEY. ✅ Confirm the connection After reconnecting, ask Claude to perform a Search Atlas action that your account supports. If the request succeeds, the MCP server is installed and authenticated. If you need further assistance, open the chat widget in the bottom-right corner of the platform and type human teammate to be connected with a member of our team.

📘 Troubleshooting MCP Connection Issues

If you are unable to connect the Search Atlas MCP or your connection is not working correctly, this guide covers the most common causes and how to resolve them. The MCP connection relies on a working login session, valid plan access, and an active authentication token. When any of these fail, the connection may not complete or tools may not work after connecting. ⚙️ Step-by-Step: Fix MCP Connection Issues Follow these checks in order to resolve the most common causes of failure. 🔍 Issue 1 — You Are Not Logged Into Search Atlas What happens When the browser window opens during connection, the login flow fails or does not complete properly. How to fix Step 1 Open: 👉 https://dashboard.searchatlas.com Step 2 Sign in to your Search Atlas account Step 3 Return to your MCP client Step 4 Retry the connection process from within your MCP client. The exact steps to retry or reconnect vary by client (for example, Claude Code, Cursor, and other MCP-compatible tools each have their own reconnection UI) — refer to your client's documentation or our connection guide for client-specific instructions. 🔍 Issue 2 — Your Plan Does Not Include MCP Access What happens - The connection may complete - But MCP tools fail when executed This is because MCP access is controlled at the plan level. How to fix Step 1 Check your Search Atlas plan in the billing area of your dashboard: 👉 https://dashboard.searchatlas.com/billing Step 2 Confirm that MCP access is included on your current plan. MCP access is available on eligible paid Search Atlas plans — if you are unsure whether your tier qualifies, review the plan details on the billing page or visit the Search Atlas pricing page. Step 3 If MCP is not included, contact your account team or upgrade your plan 🔍 Issue 3 — Session Is Expired or Stale What happens - Authentication fails - Connection stops working after setup - Tool calls return errors This occurs when the authentication token is no longer valid. How to fix Step 1 Open your MCP client Step 2 Disconnect the Search Atlas MCP Step 3 Reconnect the MCP using the current Search Atlas endpoint: https://mcp.searchatlas.com/mcp/. The Search Atlas MCP server now uses the Streamable HTTP transport (it previously used SSE), so if you have an older SSE-based URL saved in your client configuration, replace it with the Streamable HTTP endpoint above before reconnecting. The exact reconnection steps differ by MCP client (Claude Code, Cursor, and others each handle this in their own UI) — see your client's documentation if you are unsure where to update the endpoint. Step 4 Complete the login flow again This triggers a fresh authentication session. 🔍 Issue 4 — “Invalid Token” Errors What happens The client reports an invalid token even after login. Why this happens The server rejects tokens that are close to expiration instead of allowing them to fail mid-request. How to fix Step 1 Disconnect the MCP Step 2 Reconnect the MCP Step 3 Complete a fresh login 🔍 Issue 5 — Connection Works but Tools Fail What happens - MCP connects successfully - But actions fail or return errors Possible causes - Plan does not include required product area - Quota is exhausted - Permissions are restricted How to fix Step 1 Check your plan entitlements Step 2 Check your quota usage Step 3 Retry the action after confirming access 🔍 Issue 6 — MCP Tool Fails When Deploying Domain-Level Schema What happens When attempting to deploy schema at the domain level through the MCP, the tool call fails with an HTTP 400 response containing the message otto_urls not found. Why this happens This is a known platform-side issue with domain-level schema deployment via MCP. It is not caused by a misconfiguration on your side, and the fix requires a platform update. How to fix Contact Search Atlas support to confirm your account is tracked against this issue, and watch for a platform update that addresses it. In the meantime, deploy schema through the Search Atlas dashboard instead of through the MCP where possible. 🔍 Issue 7 — Platform Outage or Backend Error What happens The MCP or Atlas Brain consistently fails even after you have verified your login, plan access, and reconnected with a fresh session. This can indicate a platform-wide outage or a backend error rather than a local configuration problem. How to identify a server-side issue - The same error occurs across multiple devices - The same error occurs in an incognito or private browser window - Other Search Atlas surfaces (dashboard, API) also return errors or load slowly - The error message references a 5xx status code or a generic backend failure How to fix Check for any reported platform issues, then contact Search Atlas support with the error message, the affected tool, and the approximate time of the failure. If you have already worked through Issues 1–3 above without success, this is the recommended next step. ⚠️ Additional Notes 1. The MCP uses the same authentication and rate limits as the Search Atlas API 2. A failed tool call may return a structured error with details 3. Reconnecting the MCP resolves most session-related issues Most MCP connection issues are caused by login state, plan access, or expired sessions. By checking these in order and reconnecting when needed, you can restore full access and continue using Search Atlas through your AI assistant.

🔌 Search Atlas — MCP Server (AI Tools) Issues

This article covers: - MCP Server connectivity - Claude Desktop integration - Cursor IDE integration - MCP tool execution - SSE transport - Redis/session interruptions - tool manifest loading - API authentication - tool timeouts - and Search Atlas AI tool orchestration workflows. The Search Atlas MCP Server exposes: - 673+ tools - AI orchestration capabilities - OTTO automation - Keyword Research - Website Studio - QUEST - and AI-powered SEO workflows through a distributed MCP infrastructure. Because MCP systems depend on: - SSE connections - distributed queues - Redis - Celery workers - and upstream services …temporary interruptions, tool instability, or connection resets can occur during heavy platform activity. ⚠️ Error 1: MCP server returns "CELERY_QUEUE_NAMES not found" What's happening This is currently the most common MCP infrastructure error. The MCP deployment environment failed to load the expected Celery queue configuration during startup. When this happens: - tools may not respond - task execution may fail - or the MCP server may appear partially online. This is a server-side deployment issue. Steps to try 1. Disconnect your MCP client completely 2. Restart: - Claude Desktop - Cursor - or your MCP host application 3. Reconnect the Search Atlas MCP server 4. Wait several minutes before retrying tools 5. Refresh your MCP configuration 6. Verify your API token is still active 7. Test the connection using a lightweight tool first 8. Avoid repeatedly reconnecting every few seconds 9. Monitor whether partial tool functionality returns 10. Contact support if: - the issue persists for several hours - or ALL MCP tools remain unavailable Important Notes - this is a server-side infrastructure issue - no client-side configuration change will fully resolve it - partial recovery may occur automatically during redeployment ⚠️ Error 2: MCP connection drops — "ClosedResourceError" What's happening Search Atlas MCP uses: - SSE (Server-Sent Events) - not persistent WebSockets. Idle or long-lived SSE connections may close automatically when: - inactive too long - network conditions change - or the client sleeps/restarts. Steps to try 1. Restart your MCP client 2. Reconnect the MCP server 3. Refresh your IDE or AI application 4. Confirm your internet connection is stable 5. Avoid leaving idle MCP sessions open for long periods 6. Enable auto-reconnect behavior in your MCP client if supported 7. Verify your MCP config includes: ​"transport": "sse" 8. Update Claude Desktop or Cursor to the latest version 9. Retry the tool call after reconnecting 10. Contact support if: - SSE connections disconnect immediately - or reconnection never succeeds Important Notes - SSE disconnections are expected occasionally - well-designed MCP clients should auto-reconnect automatically ⚠️ Error 3: Redis connection error when calling MCP tools What's happening The MCP rate-limiting and session middleware temporarily lost Redis connectivity. This affects: - authentication - throttling - tool execution - and session coordination. Steps to try 1. Wait 1–2 minutes before retrying 2. Retry the SAME tool call 3. Avoid launching multiple simultaneous MCP workflows 4. Refresh your MCP session 5. Restart the MCP client if errors continue 6. Test a lightweight tool call first 7. Verify whether dashboard features still work normally 8. Monitor whether failures occur across ALL tools or only one 9. Retry during off-peak hours if instability continues 10. Contact support if: - Redis errors persist for extended periods - or all tool calls fail repeatedly Important Notes - Redis outages are usually temporary - many Redis-related failures self-recover automatically ⚠️ Error 4: "ClientDisconnect" during MCP tool execution What's happening Your MCP client disconnected before the tool response completed. This commonly affects: - full site crawls - large keyword research jobs - AI content generation - or multi-step workflows. Steps to try 1. Increase timeout settings in your MCP client 2. Break large requests into smaller operations 3. Avoid extremely large multi-tool workflows 4. Retry the failed task individually 5. Use smaller keyword batches 6. Run site audits separately from content workflows 7. Keep your MCP client window active during execution 8. Avoid sleep/hibernation while tasks are running 9. Retry during lower traffic periods 10. Contact support if: - even small tool calls disconnect consistently - or timeouts occur immediately Important Notes - long-running SEO workflows naturally require longer timeout windows - staged workflows are more reliable than massive single-pass requests ⚠️ Error 5: OTTO tool returns 500 — "An unexpected error occurred" What's happening The MCP OTTO tool could not access a valid connected project. This usually happens when: - the project is disconnected - authentication expired - or the MCP token no longer matches the account state. Steps to try 1. Go to: Left Sidebar → OTTO SEO → All Sites 2. Confirm the project is active 3. Verify OTTO is connected correctly 4. Open the project directly in the dashboard 5. Confirm the OTTO module works outside MCP 6. Go to: mcp.searchatlas.com → Settings 7. Click Regenerate Token 8. Update your MCP client configuration with the new token 9. Restart the MCP client completely 10. Contact support if: - OTTO tools fail repeatedly across all projects - or token regeneration does not restore access Important Notes - MCP tokens can become stale after account or project changes - dashboard validation is the fastest way to isolate MCP-vs-project issues ⚠️ Error 6: Keyword Research tool returns 500 from Site Explorer What's happening The MCP keyword research tools depend on upstream Site Explorer and SEO intelligence services. Temporary upstream failures may interrupt: - keyword retrieval - backlink lookups - competitor analysis - or SERP aggregation. Steps to try 1. Retry the same tool call once 2. Wait 5–10 minutes before retrying again 3. Reduce: - keyword count - countries - or competitor scope 4. Run smaller queries first 5. Test Site Explorer directly in the dashboard 6. Avoid very broad bulk requests temporarily 7. Check whether only specific regions fail 8. Refresh your MCP session 9. Retry during off-peak usage windows 10. Contact support if: - all keyword tools consistently fail - or Site Explorer itself becomes unavailable Important Notes - upstream SEO data providers occasionally throttle requests - smaller scoped queries are more reliable during high-load periods ⚠️ Error 7: Website Studio tool returns "Internal server error" What's happening The Website Studio MCP tools depend on the underlying: - Landing Page Studio - generation queues - and deployment services. If the Website Studio infrastructure is unstable, MCP tools may fail even when the connection itself remains active. Steps to try 1. Wait 10 minutes before retrying 2. Test Website Studio directly inside the dashboard 3. Confirm page generation works manually 4. Retry the MCP tool after dashboard validation 5. Reduce generation complexity 6. Break full-site requests into smaller page-level tasks 7. Refresh your MCP session 8. Restart Claude Desktop or Cursor if needed 9. Retry with smaller prompts 10. Contact support if: - Website Studio MCP fails consistently - or dashboard generation is also broken Important Notes - MCP Website Studio failures usually reflect upstream generation instability - staged generation workflows are more stable than full-site generation ⚠️ Error 8: MCP tool list is incomplete — tools are missing What's happening The MCP manifest loaded partially or failed during initialization. This can cause: - missing tools - incomplete categories - or partially registered capabilities. Steps to try 1. Disconnect your MCP client completely 2. Restart: - Claude Desktop - Cursor - or your IDE 3. Force a full manifest reload 4. In Claude Desktop: - remove the Search Atlas MCP entry - from claude_desktop_config.json 5. Re-add the MCP server configuration 6. Confirm your plan includes the missing tools 7. Wait for the manifest to finish loading fully 8. Refresh the tool list manually if supported 9. Test whether only specific categories are missing 10. Contact support if: - major tool groups never appear - or manifests repeatedly load partially Important Notes - Search Atlas exposes 673+ MCP tools - large manifests may load progressively depending on client behavior Most MCP Server and AI Tool integration issues are not caused by broken AI models. They are usually related to: - distributed queue infrastructure - SSE streaming behavior - Redis connectivity - manifest loading - upstream service instability - authentication synchronization - or long-running workflow orchestration. Because MCP systems coordinate hundreds of tools across asynchronous infrastructure, temporary instability and reconnection behavior are expected during large-scale operations. 👉 The most stable MCP workflows come from: - smaller staged tool calls - proper timeout configuration - SSE-aware clients - moderate batching - async workflow design - and responsible API usage patterns

🐛 Content Genius MCP Auto-Publishing to Wrong Site

What Is This Issue? When generating Content Genius articles through the MCP (Model Context Protocol) integration, some customers have reported that articles are automatically published live to an incorrect or unintended WordPress site — even when auto-publish settings are expected to prevent this. This behaviour appears to be specific to the MCP integration. Manual publishing and article generation through other methods do not reproduce the same problem. Is This Affecting You? You are likely experiencing this issue if: - You generated a Content Genius article using MCP. - The article published automatically to a WordPress site you did not intend. - Your auto-publish settings are configured to prevent automatic publishing. Immediate Steps to Take This is a known bug affecting Content Genius article generation via MCP. While the root cause is within the MCP integration itself, you can take the following steps to limit impact and help resolve the issue: 1. Pause MCP-based article generation for any projects connected to WordPress sites until the issue is resolved. This is the most effective way to prevent additional unintended publishing while the bug is being investigated. 2. Unpublish or delete the incorrectly published article directly from your WordPress site's admin dashboard (Pages or Posts > find the article > set to Draft or Trash) to remove it from public view immediately. 3. Review your connected WordPress sites in your Search Atlas account settings to confirm which site is set as the default publish destination for your MCP workflow, and correct it if it is pointing to the wrong site. 4. Check your auto-publish settings in Content Genius and ensure that auto-publish is explicitly disabled for any sites you do not want articles pushed to automatically. 5. Document the incident — note the WordPress site the article was published to, the site it should have been published to, the approximate date and time, your auto-publish settings, and any MCP workflow or prompt details you used. This will be needed if the steps above do not prevent recurrence. If unintended publishing continues after taking the steps above, our support team can investigate your account configuration directly. Please contact us and include the details you documented in step 5 so we can assist you as quickly as possible. How to Contact Support Reach out to our support team and include the following details to speed up the investigation: - The WordPress site the article was published to, and the site it should have been published to. - The approximate date and time the unintended publishing occurred. - Your current auto-publish settings. - Any MCP workflow or prompt details you used when the incorrect publishing occurred. Our team is actively investigating this issue. If you are experiencing unexpected publishing behaviour through the MCP integration that persists after following the steps above, please contact our support team so we can assist you directly.

🔌 Fixing MCP Tool Count Differences in Claude Code

🧭 Overview When you connect the Search Atlas MCP (Model Context Protocol) server to Claude Code, you may notice that the number of available tools is lower than the full published count. For example, you might see 475 tools instead of 607, and a specific tool or skill such as /onboard-client may appear to be missing. This is almost always expected behavior caused by client-side limits and configuration, not a fault with your account. This article explains why the difference happens and how to restore access to the tools you need. 🔎 Why the Tool Count Looks Lower The Search Atlas MCP server exposes a large catalog of tools and skills. The number that actually appears in Claude Code can be smaller for several common reasons: - Client tool limits: Claude Code caps how many MCP tools it loads at once to protect performance. When the catalog exceeds that cap, some tools are not registered. - Multiple connected servers: If you have other MCP servers connected, they share the same tool budget, leaving fewer slots for Search Atlas tools. - Permission scope: Your plan, role, or workspace permissions may hide tools you are not entitled to use. - Version mismatch: An outdated Claude Code build or MCP server version may not expose newly added tools. - Filtering or allowlists: Project or user settings can explicitly enable or disable specific tools, which changes the visible count. 💡 Why a Specific Tool Is Missing A tool like /onboard-client disappearing usually points to one of these causes: - It falls outside the loaded tool limit and was dropped when the catalog was truncated. - It is gated behind a permission or feature your current workspace does not have enabled. - It was renamed or grouped under a different skill in a newer release, so an older client no longer recognizes it. 🛠️ How to Restore Missing Tools Work through these steps in order. Most customers regain full access after the first two. 1. Update everything: Make sure Claude Code and the Search Atlas MCP connection are on their latest versions. Restart Claude Code after updating so the tool list reloads. 2. Reduce connected servers: Temporarily disconnect other MCP servers you are not using. This frees space in the tool budget so more Search Atlas tools can load. 3. Raise the tool limit: In your Claude Code configuration, increase the maximum number of MCP tools allowed if your setup permits it, then reconnect. 4. Reconnect the MCP server: Remove and re-add the Search Atlas MCP connection to force a fresh sync of the full catalog. 5. Check your permissions: Confirm your Search Atlas plan and workspace role include the tools you expect. Some tools are only available on specific tiers. 6. Search directly: If a tool is not shown in the list but you know its name, try calling it directly (for example, /onboard-client). Some clients keep tools available even when they are not displayed in the visible count. ✅ Confirming the Fix After making changes, reload Claude Code and review the tool count again. It is normal for the number to still be slightly lower than the full published total because of client limits. What matters is that the specific tools you need — including /onboard-client — are now reachable. Test the tool you were missing to confirm it works end to end. 📌 Good to Know - A tool count that is lower than the maximum is not a bug on its own; it reflects client-side limits. - Keeping only the MCP servers you actively use gives Search Atlas tools the best chance to load fully. - Tool availability can change between releases, so staying updated is the most reliable way to keep the full catalog.

🔌 Connecting Search Atlas MCP to ChatGPT

🤖 What Is the Search Atlas MCP? The Search Atlas MCP (Model Context Protocol) server lets AI assistants connect directly to your Search Atlas data. Once connected, you can ask an AI tool to pull SEO insights, run research, and work with your project data using natural language. MCP is an open standard, so any AI client that supports remote MCP servers can connect to Search Atlas — for example, the same server can also be connected to Claude, not just ChatGPT. The current server is SearchAtlas MCP v2, which is published to public MCP registries, and it authenticates using the OAuth 2.1 standard. OAuth 2.1 support is fully live for both the Claude and ChatGPT listings, so connecting is done by signing in and approving access. ✅ ChatGPT Compatibility Requirements Before you start, it is important to know which ChatGPT plans support remote MCP connectors. A common question we hear is whether the connection works on ChatGPT Plus — and the answer depends on your plan. Connector support is offered through ChatGPT's Developer Mode and custom connector features, which are not available on every plan. - ChatGPT Plus: Remote MCP connector support on the standard Plus plan is limited and may not be available in your region or build. Many connection failures happen because the account plan does not expose the custom connector option. - ChatGPT Pro, Team, Enterprise, and Edu: These plans are far more likely to include the connector and Developer Mode settings needed to add a remote MCP server. - Free plan: Custom MCP connectors are not supported. If you do not see an option to add a custom connector in your ChatGPT settings, your current plan does not support it. In that case, the issue is on the ChatGPT side, not with Search Atlas. 🔗 How to Connect the MCP Server Follow these steps to add Search Atlas as a connector in a supported ChatGPT plan. Because the Search Atlas MCP server uses OAuth 2.1, authenticating typically means signing in and approving access rather than only pasting a static key. 1. In Search Atlas, open your account settings and locate the MCP server URL and your access token or API key. 2. In ChatGPT, open Settings and find the Connectors (or Developer Mode) section. 3. Select Add custom connector or Add MCP server. 4. Paste the Search Atlas MCP server URL exactly as shown, with no extra spaces. 5. Complete authentication when prompted. With OAuth 2.1, you may be redirected to sign in to Search Atlas and authorize access; if a token is requested instead, enter your access token. 6. Save the connector and wait for ChatGPT to confirm the connection. 7. Start a new chat and confirm the Search Atlas connector appears as an available tool. 🛠️ Troubleshooting Connection Problems If the connection fails, work through these common causes in order. - Unsupported plan: Confirm your ChatGPT plan includes custom connectors. If the option is missing entirely, upgrade to a supported plan. - Incorrect server URL: Copy the MCP URL directly from Search Atlas. A missing character or trailing space will cause the connection to fail. - Invalid or expired token: Generate a fresh access token in Search Atlas and paste it again. Tokens can expire or be revoked. - Authentication errors: Make sure you are logged into the correct Search Atlas account that has access to the data you want to use. Because authentication uses OAuth 2.1, re-running the sign-in and authorization step can clear stale sessions. - Browser or cache issues: Try connecting from a different browser, clear your cache, or disable extensions that may block requests. - Feature still rolling out: Connector support is released gradually. The SearchAtlas MCP v2 server is already published to the primary public MCP registry, but the effort to submit it across every other public MCP directory is still in progress, so it may not appear in every directory search yet — adding the MCP server URL manually always works regardless of whether the server shows up in a directory listing. 💡 Tips for a Stable Connection - Keep your access token private and treat it like a password. - Reconnect after changing your password or regenerating your token. - Use clear, specific prompts so ChatGPT calls the right Search Atlas tools. - Re-test the connector after major ChatGPT or browser updates. 🆘 Still Need Help? If you have confirmed your plan supports custom connectors and the connection still fails, our team can help review your setup. If you need further assistance, click the chat icon in the bottom-right corner of the platform and type "Human Teammate" to start a live chat with a member of our team.

⚡ Understanding MCP Tool Availability in Claude Code

Overview If you notice fewer slash commands available inside Claude Code than you expected, you are not alone. This article explains why discrepancies in available MCP tools occur and what you can do if certain commands are not appearing for your account. Why You See Fewer Slash Commands Than Expected The slash commands visible inside Claude Code reflect the currently released subset of MCP (Model Context Protocol) tools for Search Atlas. The gap between what you may have seen previewed and what is live comes down to a few factors: - Phased rollout: Search Atlas releases MCP tools in stages. Not all tools have completed internal review and quality assurance at any given time. - Role and account access: Some tools are gated by account type, plan tier, or onboarding status. If your account has not completed a specific setup step, certain commands may not appear. What to Do If a Command Is Unavailable If a specific slash command is not available in your Claude Code session, here is what to check: 1. Confirm your Search Atlas account is fully set up and that onboarding steps have been completed. 2. Verify that your account plan or role includes access to the feature associated with that command. 3. If you have confirmed the above and the command is still missing, contact the Search Atlas support team for clarification on whether that tool has been released for your account type. How MCP Tool Integration Works in Search Atlas Search Atlas uses the Model Context Protocol (MCP) to expose platform capabilities as tools that AI agents like Claude Code can call directly. The tools that appear as slash commands in your session depend on which MCP tools have been released and which your account is authorized to access. If you need help confirming your current tool access or understanding which commands should be available for your account, contact the Search Atlas support team for assistance.

📘 MCP Best Practices (Unified)

MCP is powerful because it connects natural language to real platform actions — but that also means every request can trigger real execution, quota usage, or external impact. This guide consolidates best practices across all MCP usage patterns to help you work efficiently, avoid errors, and maintain full control over your actions. 🧠 Core Principle 👉 MCP is most effective when you separate analysis from execution. - First: retrieve and analyze data (read-only) - Then: execute actions (work-producing) This reduces errors, prevents wasted credits, and improves outcomes. ⚙️ Step-by-Step: The Ideal MCP Workflow Step 1 — Start with Read-Only Actions Ask your assistant to retrieve data first. Examples: - “Show my SEO audit results” - “Check my backlinks” - “Show my GBP locations” 👉 These actions do NOT consume Search Atlas quota. Step 2 — Analyze Before Acting Review the data returned: - identify issues - compare performance - understand context 👉 This step ensures your next action is intentional and accurate. Step 3 — Ask for Recommendations Before executing, ask: - “What should I do next?” - “What are the best optimizations?” 👉 This helps avoid unnecessary or incorrect actions. Step 4 — Execute Actions Once ready, ask for execution: - “Apply fixes” - “Generate content” - “Create campaign” 👉 These actions consume quota because they produce work. Step 5 — Review Before Publishing or Submitting Before any external or irreversible action: - review content - validate data - confirm intent Step 6 — Approve Sensitive Actions If prompted: - review carefully - approve only if correct 👉 Required for: - spending actions - destructive actions - external integrations Step 7 — Monitor Results After execution: - review outputs - check performance - refine next steps 🔁 The Golden Workflow 👉 Read → Analyze → Recommend → Execute → Review This is the most reliable way to use MCP across all product areas. ⚖️ Managing Quota & Costs Understand When Quota Is Used 👉 Quota is charged when work is produced Examples: - generating content → consumes quota - running audits → consumes quota - viewing data → does NOT consume quota Step-by-Step: Avoid Unnecessary Usage Step 1 Ask: - “Will this consume credits?” Step 2 Check quota: - “Show my usage” Step 3 Proceed only if needed Best Practice 💡 Always validate before executing large or batch actions. 🔐 Managing Risk & Control Use Approval as a Safety Layer Approval exists to prevent: - unintended spending - destructive changes - irreversible actions 👉 Never approve blindly — always review first. Separate Preparation from Execution Instead of: ❌ “Create and publish content immediately” Use: ✅ “Create content” → review → “Publish content” Be Aware of External Impact Integration actions may: - publish publicly - trigger external systems - consume credits or HDC 🧠 Writing Better Prompts Be Specific Instead of: ❌ “Improve my SEO” Use: ✅ “Run a site audit and show top issues” Break Down Complex Tasks Instead of: ❌ “Do everything for my SEO” Use: 1. “Run audit” 2. “Show recommendations” 3. “Apply fixes” Iterate Refine based on results: - “Improve this section” - “Optimize further” ⚠️ Avoiding Common Mistakes Mistake 1 — Executing Without Reviewing Fix: Always analyze first Mistake 2 — Ignoring Quota Fix: Check usage before large actions Mistake 3 — Approving Without Reading Fix: Treat approval as a checkpoint Mistake 4 — Vague Requests Fix: Use clear, specific instructions Mistake 5 — Running Too Many Actions Quickly Fix: Space requests to avoid rate limits 🛠️ Efficiency Tips Use Read-Only for Exploration No quota cost Batch Thoughtfully Group related actions, but review before execution Use MCP Iteratively Work step-by-step instead of all at once Monitor Usage Regularly Avoid hitting limits mid-workflow 🔁 How This Applies Across All Product Areas These best practices apply to: - SEO (OTTO) - Content (Content Genius) - GBP (Local SEO) - PPC - PR & Social - AI Visibility (formerly LLM Visibility) - Integrations 👉 The workflow stays the same — only the tools change. 🧠 Key Takeaways - MCP is most powerful when used step-by-step - Separate analysis from execution - Quota is only used when work is produced - Approval protects high-impact actions - Clear prompts improve results MCP is designed to give you full control over Search Atlas through conversation. By following a structured workflow — starting with data, validating decisions, and executing intentionally — you can maximize efficiency, minimize errors, and confidently use the full power of the platform through your AI assistant.

📘 MCP for Content Creation (Content Genius)

📘 MCP for Content Creation (Content Genius) Use MCP to generate, manage, and publish SEO content using the Content Genius product area. Content Genius is the Search Atlas system for AI-powered article generation, brand voice management, and content publishing workflows. Through MCP, your AI assistant can execute the full content lifecycle on your behalf. 🧠 What You Can Do Using MCP for content creation, you can: - Generate SEO articles from keywords or topics - Auto-generate content using the Deep Knowledge Network (DKN) - Manage your brand voice through Brand Vault - Edit, review, and regenerate content - Publish content to WordPress, GHL (GoHighLevel), and other CMS connectors - Schedule content in batches - Export content to Google Docs ⚙️ Step-by-Step: Generate an Article Step 1 Open your MCP-connected AI assistant Step 2 Ask for content generation. Examples: - “Write an SEO article about [keyword]” - “Generate a blog post for my content project” Step 3 The assistant generates the article. The content is created using AI based on your request and available data. Step 4 Review the draft. You can read, edit, or refine the content before publishing. 🔁 Step-by-Step: Edit or Regenerate Content Step 1 Ask your assistant: - “Edit this article to improve SEO” - “Rewrite this section” - “Regenerate this content” Step 2 The assistant updates the content Step 3 Review the updated version. You can repeat this process until the content meets your requirements. 🧱 Step-by-Step: Use Brand Voice (Brand Vault) Step 1 Ensure your brand information is set in Brand Vault Step 2 Ask your assistant: “Generate content using my brand voice” Step 3 The assistant applies your brand voice. Content is generated using your stored business information, tone, and sources. 🚀 Step-by-Step: Publish Content Step 1 Ask your assistant: - “Publish this article to WordPress” - “Post this content to my CMS” (e.g., GHL) Step 2 The assistant prepares the publication Step 3 Confirm the action (manual approval is required) Step 4 The article is published through the connected platform ✅ Approval Requirement: Articles generated through Content Genius require explicit manual approval before they are published to any connected CMS (WordPress, GHL/GoHighLevel, etc.). Reconnecting a CMS connector does not bypass this approval step — content will remain in a pending state until you approve it. 📅 Step-by-Step: Schedule Content Step 1 Ask your assistant: - “Schedule this article for next week” - “Create a batch of scheduled posts” Step 2 The assistant sets the schedule Step 3 Content is queued for future publishing 📤 Step-by-Step: Export Content Step 1 Ask your assistant: “Export this article to Google Docs” Step 2 The assistant prepares the export Step 3 The content is delivered in the selected format ⚠️ Important Note - AI-generated content consumes Search Atlas quota - Publishing and scheduling already-generated content does not consume additional quota - Some actions may require confirmation before execution - Publishing always requires manual approval before content goes live on a connected CMS 🔁 How This Fits Together The content workflow through MCP typically follows: 1. Generate content 2. Review and edit 3. Apply brand voice 4. Approve, then publish or schedule 5. Export if needed All steps are executed through your AI assistant using the same tools available in the Search Atlas platform. MCP allows you to manage your entire content production workflow through conversation — from generating articles to publishing and scheduling them. 🛠️ Troubleshooting GHL Connector Auto-Publishing Without Approval A previously reported issue (internal reference: Linear ticket SPE-505) caused Content Genius articles to be automatically published to GHL (GoHighLevel) blogs without manual approval after the GHL connector was reconnected. This bug has been resolved by the engineering team and the fix has been applied across all affected accounts — no action is required on your end. If you continue to observe articles publishing to GHL or any other CMS without manual approval, please contact Search Atlas support so the issue can be escalated.

📘 MCP for LinkLab

LinkLab is a publisher marketplace that allows you to browse publications, create content for placement, and complete orders. Through MCP, your AI assistant can manage the full workflow from research to checkout and tracking. This article explains how to find publications, create placements, and manage LinkLab orders using MCP. In LinkLab, a placement is a paid content slot on a third-party publication where your article will be published as a guest post or sponsored content. 🧠 What You Can Do Using MCP with LinkLab, you can: - Browse available publications - View publication details - Create articles for placement - Add placements to a cart - Complete checkout - Track order and publication status - Handle rejected publications using the AlternativesDrawer (Choose for Me / Refund HDC) 👉 LinkLab actions may involve Hyperdrive Credits (HDC) for paid placements. ⚙️ Step-by-Step: Browse Publications Step 1 Open your MCP-connected AI assistant Step 2 Ask to view available publications Examples: - “Show available publications” - “Find sites for article placements” Step 3 The assistant retrieves publication listings Step 4 Review publication details This may include: - publication information - placement opportunities 🔍 Step-by-Step: View Publication Details Step 1 Ask your assistant: - “Show details for this publication” Step 2 The assistant retrieves publication data Step 3 Review: - domain authority - pricing - niche / topical category - placement types available - publication characteristics ✍️ Step-by-Step: Create an Article for Placement Step 1 Ask your assistant to create an article and provide inputs such as a target keyword, topic, target URL, and desired word count. You can either ask MCP to generate the article from scratch or provide an existing draft for it to refine. - “Create an article for this placement targeting the keyword ‘best CRM software,’ around 800 words, linking to example.com/crm” Step 2 The assistant generates the article from the inputs you provided, or prepares the draft you supplied for placement Step 3 Review and refine content 👉 Articles are prepared specifically for placement within publications. 🛒 Step-by-Step: Add to Cart Step 1 Ask your assistant: - “Add this placement to my cart” Step 2 The assistant adds the item Step 3 Review cart contents 💳 Step-by-Step: Checkout and Purchase Step 1 Ask your assistant: - “Checkout my LinkLab cart” Step 2 The assistant prepares the checkout Step 3 Review cost and details Step 4 Approve the purchase 👉 Checkout actions involve Hyperdrive Credits (HDC) and require confirmation. Step 5 The order is completed 📊 Step-by-Step: Track Orders Step 1 Ask your assistant: - “Show my LinkLab orders” Step 2 The assistant retrieves order data Step 3 Review: - order status - publication progress - placement updates 👉 Order edit limitations: LinkLab orders are generally not directly editable after submission (for example, adding reference links or removing placeholder content). If you need a post-submission change, contact support or the fulfillment team to request the update. 🔄 Rejected Publications & the AlternativesDrawer A publication may be rejected when the publisher declines the submitted article (for example, due to editorial guidelines, niche mismatch, or content policy). When this happens, the AlternativesDrawer opens to help you resolve the placement. The AlternativesDrawer offers two one-click actions: - Choose for Me — automatically selects a comparable alternative publication and re-routes your placement to it. - Refund HDC — returns the Hyperdrive Credits used for the rejected placement back to your balance. You can trigger either action directly in the AlternativesDrawer in the UI, or by asking your MCP-connected assistant — for example, “Choose an alternative publication for my rejected placement” or “Refund HDC for the rejected placement.” 💰 HDC & Credits - When HDC are deducted: Hyperdrive Credits are charged at checkout, when the placement order is confirmed. - If a publication is rejected: the credits used for that placement become eligible for refund. You can recover them via the Refund HDC action in the AlternativesDrawer, or choose an alternative publication with Choose for Me to apply the credits to a new placement instead. - How to request a refund: Use the Refund HDC button in the AlternativesDrawer, or ask your MCP assistant (“Refund HDC for this rejected placement”). If the refund does not appear in your balance, contact support. 🛠️ Troubleshooting “Tool execution not approved” error This message means the MCP action requires your explicit approval before it can run. MCP requests confirmation for actions that affect your account — such as adding items to a cart, checking out, or spending HDC. To resolve it, approve the tool call in your AI assistant interface when the approval prompt appears (for example, click Approve or Allow on the pending MCP action). If you continue to see the error after approving, retry the request, and contact support if the issue persists. Content Genius → LinkLab bridge issues When pushing articles from Content Genius directly into a LinkLab placement, the publishing bridge may occasionally return a blocked pipeline state. If your article does not transfer or the pipeline stalls, contact support so the team can unblock the pipeline and complete the placement. ⚠️ Important Notes 1. LinkLab placements are paid actions 2. Checkout consumes Hyperdrive Credits (HDC) 3. Purchases require approval before execution 4. Orders involve external publication processes 5. Orders are generally not editable after submission — contact support for post-submission changes 6. Rejected publications can be resolved with Choose for Me or Refund HDC in the AlternativesDrawer 🔁 How This Fits Into MCP Workflows A typical workflow: 1. Browse publications 2. Select placement opportunities 3. Create article 4. Add to cart 5. Checkout 6. Track order 7. Resolve rejections via the AlternativesDrawer if needed 👉 This supports end-to-end link placement management. 🧠 Best Practices - Review publication details before purchasing - Ensure article content is accurate and relevant - Confirm costs before checkout - Monitor order status after purchase - If a publication is rejected, act quickly in the AlternativesDrawer to either re-route the placement or refund the HDC MCP for LinkLab allows you to manage article placements through conversation. From discovering publications to completing purchases and tracking results, your AI assistant can execute each step using the same capabilities available in Search Atlas, including paid placement workflows powered by Hyperdrive Credits.

📘 Connect MCP in Claude (Desktop & Browser)

The Search Atlas MCP can be connected through Claude's Custom Connector experience. Both Claude Desktop and Claude.ai use the same MCP endpoint and authentication flow. This article walks through the connection process and explains what to expect during setup. ⚙️ Step-by-Step: Connect MCP in Claude ⚙️ Step-by-Step: Connect MCP in Claude Step 1 Open Customize in Claude. Step 2 Select Connectors. Step 3 Click Add connector. Step 4 Choose Add custom connector. Step 5 Name the connector Search Atlas. Step 6 Paste the Search Atlas MCP server URL: https://mcp.searchatlas.com/mcp/ Step 7 Click Add. Step 8 Click Connect. Step 9 Claude opens your Search Atlas account so you can grant access. Step 10 Click Authorize. Step 11 After authorization, the connector is connected and ready to use in Claude. Step 12 To verify the connection, open a Claude chat and ask: "Are you connected to my Search Atlas account?" 🔐 What Happens After You Connect Once connected, Search Atlas tools become available directly inside Claude chats. You can ask Claude to confirm the connection by opening a chat and asking, "Are you connected to my Search Atlas account?" After the connection is confirmed, you can start using your Search Atlas account through Claude. From then on, the client holds and refreshes the credential automatically. That is the supported OAuth path for customer-facing scenarios. Those actions continue to follow the same account authentication, plan entitlements, and quota rules used in the Search Atlas app. ⚠️ Troubleshooting If the connection does not complete, start with these checks from the source: Check 1 — Are you already logged into Search Atlas? If not, sign in to dashboard.searchatlas.com first, then retry the connection flow. Check 2 — Does your plan include MCP access? If MCP is not included in the plan, the OAuth flow may complete but tool calls will fail with an authorization error. Check 3 — Is the session stale or expired? Disconnect the MCP in Claude and reconnect to trigger a fresh login. Oce you complete the Claude connection flow, Claude becomes a conversational interface for the Search Atlas tools available in your account. You describe the outcome, and Claude can execute the corresponding Search Atlas actions through the MCP using your existing permissions, entitlements, and quotas.