MCP for AI agents

Connect AI coding agents (Claude Code, Codex, Cursor, VS Code, Claude.ai and any other client that speaks the Model Context Protocol) to your analytics. Ask them about your numbers, or have them build what you need: a signup funnel, a retention insight, a dashboard for the launch, conversion goals.

Connect an agent#

The MCP server lives at https://api.anyanalytics.org/mcp. Agents sign in with OAuth: the first time, a browser tab opens where you choose what the agent may do.

Claude Code

claude mcp add --transport http anyanalytics https://api.anyanalytics.org/mcp

Then run /mcp in Claude Code, pick anyanalytics and Authenticate.

Codex

codex mcp add anyanalytics --url https://api.anyanalytics.org/mcp
codex mcp login anyanalytics

Cursor

{
  "mcpServers": {
    "anyanalytics": {
      "url": "https://api.anyanalytics.org/mcp"
    }
  }
}

Cursor shows Needs login next to the server: click it to authorize.

VS Code

{
  "servers": {
    "anyanalytics": {
      "type": "http",
      "url": "https://api.anyanalytics.org/mcp"
    }
  }
}

Claude.ai and Claude Desktop

Open Settings → Connectors → Add custom connector, paste https://api.anyanalytics.org/mcp and click Connect.

Choosing what an agent may do#

When an agent connects, you pick:

1. The organization

   Its sites are the only ones the agent can see.

2. An access level

   Level	The agent can
   Read only	Read reports, saved insights, dashboards and the list of events your sites receive.
   Can build	Also create and edit funnels, insights, dashboards and conversion goals.
   Full control	Also delete them.

3. Which sites

   All sites (including ones you add later), or a list of sites.

An agent acts as you and never gets more than your own role allows. As a member (not an owner or admin), your agents can change only what you created and can't manage goals. If you leave the organization, your agents lose access too. See Team and roles.

Managing connections#

Every connection is listed under MCP in your organization's sidebar, with its access level, its sites, when it was last used and how many calls it made. From there you can:

• Activity: see every tool call (what, on which site, whether it worked), kept 90 days.
• Edit access: change the level or the sites. It applies from the agent's next request.
• Disconnect: its tokens stop working at once.

Owners and admins see everyone's connections. They can lower another member's access or disconnect it, but only the person who connected an agent can give it more.

Funnels, insights, dashboards and goals an agent creates show a badge with the agent's name, so you always know where they came from.

What agents can do#

Every answer includes links to open the result in the dashboard. Queries use the same shapes as the Read API.

Only real events

Before an agent saves a funnel, an insight or an event goal, the events it uses are checked against what your site received in the last 90 days. A typo or an invented event is refused with suggestions, like No events named "sign_up" (did you mean "signed_up"?). For tracking you are shipping right now, the agent can say so explicitly (allowUnseenEvents).

Access tokens#

For clients without OAuth sign-in (CI jobs, scripts), create a personal access token under MCP → Connect another client → Access token. It starts with vxm_, is shown once, acts as you with the access you chose, and works until you disconnect it.

claude mcp add --transport http anyanalytics https://api.anyanalytics.org/mcp \
  --header "Authorization: Bearer vxm_YOUR_TOKEN"

Limits#

• 300 requests per minute and 8 at once per connection.
• Report results are cached for a minute, like in the dashboard.
• If your free trial has ended, report tools answer with an error until you choose a plan.

Looking for the report on AI crawlers visiting your site instead? That's AI agents.