Connect your AI (MCP)

VisitorType ships a built-in Model Context Protocol server, so you can manage everything from your preferred AI — Claude, ChatGPT, Cursor, or any MCP-compatible client. Ask your assistant which AI agents visited this week, have it create triggers and tags, and let it publish — all against your account.

1. Create an API key

Go to Dashboard → Agents and create a key. It's shown once — store it like a password. Revoke keys anytime.

2. Add the server to your client

Claude Code:

claude mcp add --transport http visitortype https://visitortype.ai/api/mcp \
  --header "Authorization: Bearer aitm_YOUR_KEY"

Claude Desktop, Cursor, and most other clients (JSON config):

{
  "mcpServers": {
    "visitortype": {
      "type": "http",
      "url": "https://visitortype.ai/api/mcp",
      "headers": { "Authorization": "Bearer aitm_YOUR_KEY" }
    }
  }
}

3. Available tools

Tool	What it does
`list_containers`	Your containers with tag/event counts
`get_container`	Full setup: tags, triggers, published version
`get_analytics`	Events by visitor type, top AI agents, top paths (1–90 days)
`list_recent_events`	Latest visits, filterable by visitor type
`classify_user_agent`	Run any UA string through the detection engine
`create_trigger`	New trigger from a list of conditions
`create_tag`	New tag (custom_html, image_pixel, data_layer), optionally attached to a trigger
`publish_version`	Freeze and ship the current setup

Things to ask

“Which AI agents hit my site this week and what did they read?” · “Create a trigger for Perplexity visitors and a tag that points them to /llms.txt, then publish.” · “Is this user agent one of the known AI crawlers:Mozilla/5.0 … GPTBot/1.2?”

Writes are real: create_tag and publish_version change what your site serves. Give keys only to clients you trust, and revoke from the Agents page if in doubt.