MCP Server

Every question, deploy annotation, and site metric, served through a single authenticated Model Context Protocol connection.

Connection details

Server URL	`https://api.staats.ai/functions/v1/mcp/YOUR_ACCOUNT_KEY`
Transport	Streamable HTTP (JSON-RPC 2.0 over POST)
Authentication	Key in the URL path (zero headers needed)

Your account key starts with ak_ and lives in your dashboard. Because the key is embedded in the URL, native remote MCP configuration takes one line:

# Claude Code CLI (--scope user connects every project on the machine)
claude mcp add --scope user --transport http analytics "https://api.staats.ai/functions/v1/mcp/YOUR_ACCOUNT_KEY"

# Codex
codex mcp add analytics https://api.staats.ai/functions/v1/mcp/YOUR_ACCOUNT_KEY

// Cursor (~/.cursor/mcp.json or project .cursor/mcp.json)
{
  "mcpServers": {
    "analytics": { "url": "https://api.staats.ai/functions/v1/mcp/YOUR_ACCOUNT_KEY" }
  }
}

The URL acts as your credential: keep it out of public repositories. Clients that prefer standard header authorization can send Authorization: Bearer YOUR_ACCOUNT_KEY to https://api.staats.ai/functions/v1/mcp directly. Leaked a key? Regenerate it from the dashboard — the old key stops working within a minute, and your sites' script tags are unaffected.

Site selection semantics

If your account has a single site, all tools execute against it automatically. When multiple sites exist, tools take an optional site argument: the public data-site code from the site's script tag (also returned by list_sites alongside each domain and name). Exact matches only; the server never guesses.

Available tools

Tool	Purpose & Behavior
`get_overview`	Summary of visitors, pageviews, top paths, referrers, and trend vs previous period.
`query`	Aggregated metrics by dimension with flexible filtering (paths, referrers, countries, events).
`funnel`	Ordered conversion steps (paths or events) with stage-by-stage drop-off calculations.
`what_changed`	Automatic anomaly detection: traffic surges, viral referrers, and unusual days.
`record_annotation`	Record deploy markers with commit hash and touched routes for before/after measurement.
`list_annotations`	Retrieve chronological change logs and deploy milestones.
`compare_around`	Compare metrics before vs after a deploy (with route-scoped analysis when routes were logged).
`get_site_context` / `update_site_context`	Persistent site memory across sessions: site goals, audience, event dictionaries, and open notes.
`list_sites`	List all sites on the account with public site codes.
`create_site`	Create a new site and retrieve its script tag without leaving your agent chat.
`delete_site`	Permanently delete a site and its data (requires typed domain confirmation).
`portfolio_overview`	Multi-site portfolio breakdown sorted by highest traffic and growth.

Tool responses are returned as token-efficient JSON structured for LLM reasoning. Proactive quota notices are returned from 80% usage onward.