Skip to content
  • There are no suggestions because the search field is empty.

Does OtterlyAI have an MCP?

Yes, OtterlyAI provides an MCP server that lets you query your AI search visibility data directly from your AI assistant, without leaving your workflow.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants like Claude, Cursor, and other compatible clients call external tools and live data sources in a structured way.

When connected, your AI client can request data from OtterlyAI on your behalf and use it to answer questions directly — no manual exporting, copy-pasting, or context-switching required.

How can the OtterlyAI MCP help me?

Access your AI search visibility data directly in your AI tool

Ask questions about brand coverage, prompt rankings, citation performance, or competitor visibility. Your AI assistant retrieves the latest OtterlyAI data and includes it in its response, no need to open the dashboard or export files.

Get AI-generated recommendations without leaving your workflow

Ask Claude, ChatGPT or Cursor what OtterlyAI recommends for improving your brand's AI search visibility. The assistant fetches the latest recommendations for your brand report and surfaces them instantly.

How does it work?

The OtterlyAI MCP server is available at https://data.otterly.ai/mcp. It uses the MCP streamable HTTP transport and authenticates entirely via OAuth 2.0, you do not need to pass an API key. When you add the server to a compatible client, the client walks you through a standard browser sign-in to OtterlyAI.

Compatible clients include Claude.ai, Claude Code, n8n, ChatGPT, Claude Desktop, Cursor, and any other client that supports the MCP streamable HTTP transport.

For step-by-step connection instructions for each client, see the official MCP documentation →

For connection to MS Copilot - check out Microsoft documentation. We use dynamic OAuth 2.0.

Video walkthroughs

If you prefer to follow along visually, we've put together short setup guides:

📺 How to connect OtterlyAI MCP to Claude



📺 How to connect OtterlyAI MCP to ChatGPT

What OtterlyAI data does the MCP provide?

Brand visibility & KPIs

  • Brand coverage history and rank across AI engines, with sentiment
  • Prompt-level brand mention stats
  • Engine breakdown (ChatGPT, Google, Perplexity, Copilot, Google AI Mode, Gemini, Claude)
  • Top domains per brand report

Prompts & AI responses

  • Full list of prompts tracked for your brand report, with brand and domain mention stats
  • Raw AI engine responses, brand mentions, and citations
  • Prompt filtering by engine, tag, date range, country, and sort order
  • Create and delete prompts, and create and delete tags — manage your prompt set and tags directly from your AI assistant (requires write permission)

Citations

  • Cited URLs for your brand, filterable by AI engine, tag, and search query
  • Share-of-citations and top cited domains over any date window
  • Which prompts cited a specific URL

Recommendations

  • AI-generated recommendations for improving brand visibility
  • Filterable by engine and country

GEO audits

  • Crawlability checks, robots.txt analysis and bot access per crawler (GPTBot, ClaudeBot, PerplexityBot, and more)
  • Content / AI-readiness checks, AI readiness score, structural and dynamic-content scores
  • On-demand audit triggers for any URL or runs a query fan out (requires write permission)

Available categories

The MCP server covers four tool categories: Workspaces & engines, Brand reports, GEO audits, and Write tools (write permission required). For the full reference including required arguments, optional filters, and description, see the official MCP documentation →

Frequently asked questions

What does a 401 or 403 error mean?

A 401 means your OAuth token is missing or expired, reconnect the MCP server to refresh it. A 403 means your account lacks permission for the requested workspace.

Why does the authorization tab keep reopening, or it says it can't connect?

This usually happens when two MCP clients on the same machine (for example Claude Desktop and Cursor, or two Claude Desktop profiles) both use the mcp-remote bridge pointed at the same server URL. They share one stored login, so whichever client authorizes last knocks the other offline — which then reopens its own authorization tab and knocks the first one out. Reconnecting fixes one and breaks the other, so the loop never ends on its own. To fix it, give each client its own credential store and callback port. See Running more than one client in the MCP documentation for the exact config.

Can I run OtterlyAI's MCP in more than one tool at the same time (e.g. Claude Desktop and Cursor)?

Yes, but if both use the mcp-remote bridge you need to point each one at a separate credential store and a separate callback port, otherwise they'll share a single login and conflict. Note that Claude.ai, Claude Code, and ChatGPT each hold their own token natively, so they don't run into this — the conflict is specific to mcp-remote-based setups. Full setup instructions are in the MCP documentation.

Can I connect two different OtterlyAI accounts on the same machine?

Yes, but the same rule applies: each client needs its own credential store and callback port. If two clients share one store, both end up on whichever account authorized last. Using a separate browser profile per account is the reliable way to keep them apart.

I got a "state already in use" error while connecting. What do I do?

A connection link is single-use, and retrying the same one triggers this. Wait about ten minutes, then start the connection again from the client itself rather than reloading the browser page. See here.

The wrong OtterlyAI account got connected. How do I switch it?

The connection is approved by whichever account your browser is already signed into, the client doesn't choose. Sign the account you want into the browser first (a separate browser profile per account is the most reliable approach), then authorize the connection.

Do I need an API key to use the MCP server?

No. The MCP server authenticates via OAuth 2.0 only. Your API keys are for the REST Public API and the Claude skill only, the MCP server does not accept them.

Where can I check my usage and how many requests I have left?

You can check this under Plan usage in your Billing tab.

Where can I check the limits for each plan?

You’ll see how many MCP requests are available per plan on our Pricing page, you can also see this information in your Billing tab.

I’m on a trial period, do I also get an MCP limit?

Yes, you’ll get limits to test the MCP.

What are some available tools I can connect to?

You can connect to Claude, n8n, Cursor, ChatGPT and many more. Connecting is available with any tool that supports the MCP streamable HTTP transport.

Related articles