Glai

Docs

Ask your AI assistant about your traffic

Connect Claude, ChatGPT, Cursor or VS Code to Glai and ask in plain words: where did visitors come from this week, which pages bring paying customers, who is on the site right now. The assistant reads your numbers through the Glai MCP server, after you sign in or with an API key, and it can only read.

Connect

By signing in: Claude and ChatGPT

Nothing to paste but the address. Owners and admins of a workspace on a paid plan can connect it.

  1. Claude (web and desktop): Settings → Connectors → Add custom connector. ChatGPT: Settings → Apps and connectors, with developer mode on → Create. Give it the name Glai and the address https://mcp.glai.io/mcp.
  2. Press Connect. Glai opens: sign in, pick the workspace and press Allow.
  3. Ask "Which Glai sites can you see?" to check it works. To disconnect, use Settings → API keys → Connected apps.

With an API key: Claude Code, Cursor, VS Code

  1. Create a key under Settings → API keys (owners and admins, on a paid plan). Give it a name like "Claude" so you can revoke it on its own later.
  2. Add the server to your assistant with the address https://mcp.glai.io/mcp and the header Authorization: Bearer gl_live_…, as below.
  3. Ask "Which Glai sites can you see?" to check it works.

Claude Code

Request
claude mcp add --transport http glai https://mcp.glai.io/mcp \
  --header "Authorization: Bearer gl_live_your_key"

Claude Desktop with a key

Signing in (above) is simpler. With a key instead: Settings → Developer → Edit config, add this to claude_desktop_config.json and restart Claude. It needs Node.js installed.

claude_desktop_config.json
{
  "mcpServers": {
    "glai": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.glai.io/mcp", "--header", "Authorization:${GLAI_AUTH}"],
      "env": { "GLAI_AUTH": "Bearer gl_live_your_key" }
    }
  }
}

Cursor

~/.cursor/mcp.json
{
  "mcpServers": {
    "glai": {
      "url": "https://mcp.glai.io/mcp",
      "headers": { "Authorization": "Bearer gl_live_your_key" }
    }
  }
}

VS Code

.vscode/mcp.json
{
  "servers": {
    "glai": {
      "type": "http",
      "url": "https://mcp.glai.io/mcp",
      "headers": { "Authorization": "Bearer gl_live_your_key" }
    }
  }
}

What it can do

Ten tools, every one of them read only. Each calls one route of the API, so the assistant sees exactly what the key could already read.

ToolWhat it doesAPI route
whoamiWhich workspace and plan the key belongs to. The assistant calls it first./api/v1/me
list_sitesYour sites with their ids and who is on each now./api/v1/sites
get_statsVisitors, visits, pageviews, bounce rate, time on site, revenue and conversion for a range, against the period before./api/v1/sites/{site}/stats
get_breakdownTop sources, pages, countries, devices, campaigns, AI assistants, events and more, with revenue where it applies./api/v1/sites/{site}/stats
get_timeseriesThe numbers per hour, day or month, with the previous period beside them./api/v1/sites/{site}/stats
get_liveWho is on the site right now and what they are reading./api/v1/sites/{site}/live
list_visitorsRecent visitors, paying or identified ones, or a search by email./api/v1/sites/{site}/visitors
get_visitorOne visitor's journey, visit by visit, and their payments./api/v1/sites/{site}/visitors/{id}
list_paymentsPayments with the source and campaign that brought each customer./api/v1/sites/{site}/payments
get_goalsGoals and their conversions./api/v1/sites/{site}/goals

Things to ask

  • How many visitors did my site get this week compared with last week?
  • Which sources brought paying customers this month, and how much revenue each?
  • How much traffic comes from ChatGPT, Perplexity and other AI assistants?
  • Which pages do people land on before they sign up?
  • Who is on the site right now, and what are they reading?
  • Show me the journey of my last paying customer.

Your data

  • Read only. No tool changes a site, a goal, a setting or a payment.
  • Nothing is stored on the MCP server. It holds no key, database or history: your key or sign-in comes with each request, is passed to the Glai API and forgotten. Its logs keep a short fingerprint, never the key or what was asked.
  • A sign-in reads one workspace. The one you picked when you pressed Allow. It stops working if you leave that workspace, stop being its owner or admin, or the plan no longer includes the API, and you are asked to approve it again after a year.
  • What the assistant reads goes to its maker. Numbers, pages and, for identified visitors, their email and name are sent to the assistant you connected (Anthropic, Cursor, Microsoft…) under that company's terms. Leave identified visitors out by not asking about them, or do not connect a workspace whose visitors you identify.
  • Revoke at any time. Revoking the key, or pressing Disconnect under Connected apps in Settings → API keys, cuts the assistant off at once. The API's limit of 300 requests a minute per workspace applies.