October 1, 2026

MCP server: AEOTrack in Claude, Cursor and other assistants (early access)

What it does

Early access. The AEOTrack MCP server lets Claude Desktop, Claude Code, Cursor, VS Code or any other client that speaks the Model Context Protocol ask your account questions in plain language — "how visible is acmecoffee.example in AI answers?", "which sites does ChatGPT cite instead of us?", "track 'best coffee beans for cold brew' and check it now". It reads the same numbers as the dashboard and the public API, with your own API key, and holds no data of its own.

It is on the Agency plan only. A key from a lower plan can connect and list the tools, but every tool returns an error naming the plan it needs, with no data and nothing spent. Early access means the tool set and the setup may still change; tell support@aeotrack.io what works and what doesn't.

The seven tools

Tool	What it does	Spends
`list_domains`	Websites on the account, with their `domain_id`	–
`get_score(domain_id)`	Questions score and Topics score, mention rate, ± range, engine coverage	–
`list_questions(domain_id, include_snippets=false)`	Every tracked question with per-engine verdicts	–
`add_question(domain_id, question)`	Track a new question (the plan's question limit applies)	–
`run_check(domain_id, query_id)`	Ask every engine now; 24-hour cooldown per question	One on-demand re-check from the monthly pool (the first check of a question is free)
`get_sources(domain_id)`	Most-cited source domains and your website's own cited URLs	–
`get_share_of_voice(domain_id)`	Mentions against named competitors, last 30 days	–

Your API key

Create an API key under Settings → Company → API keys (owners and admins; the key is shown once and starts with aeo_live_ — see The public API and API keys). Give it a name such as MCP server so you can revoke it on its own.

Connect without installing (Cursor, VS Code, Claude Code)

AEOTrack hosts the server. Point your client at the endpoint and send the key as a bearer token:

Endpoint: https://api.aeotrack.io/mcp
Header: Authorization: Bearer aeo_live_…

Cursor and VS Code: put this in .cursor/mcp.json (or the global file under Cursor → Settings → MCP), or in your VS Code MCP settings:

{
  "mcpServers": {
    "aeotrack": {
      "url": "https://api.aeotrack.io/mcp",
      "headers": {
        "Authorization": "Bearer aeo_live_…"
      }
    }
  }
}

Claude Code: one command:

claude mcp add --transport http aeotrack https://api.aeotrack.io/mcp --header "Authorization: Bearer aeo_live_…"

Claude Desktop

Claude Desktop runs MCP servers as local commands, so it uses the small aeotrack-mcp package, which talks to the same API. Install it with pipx (Python 3.10 or later):

pipx install aeotrack-mcp

Run aeotrack-mcp once to check it starts: it waits on stdin, which is the stdio transport working; Ctrl-C to stop. Then add it to claude_desktop_config.json (Claude Desktop → Settings → Developer → Edit Config) and restart Claude:

{
  "mcpServers": {
    "aeotrack": {
      "command": "aeotrack-mcp",
      "env": {
        "AEOTRACK_API_KEY": "aeo_live_…",
        "AEOTRACK_API_URL": "https://api.aeotrack.io"
      }
    }
  }
}

The package checks your plan before its first call, so a key below Agency gets the plan message straight away.

Things to know

The key is sent as Authorization: Bearer to the hosted endpoint, or as X-API-Key by the package. Never paste it into a chat with an assistant.
run_check asks all the engines for real and can take one to two minutes.
Rate limit: 120 API calls per minute per key, the same as the public API. A chat that loops over every question on every site will hit it — ask for the summary first.
Revoking the key under Settings stops the server on its next call.