Ask your analytics

Lynq is an MCP server. Claude Code, Cursor, Codex and any other client that speaks the Model Context Protocol connect to it with one key and ask the same questions the dashboard answers: how the site is doing, which pages hold people, whether Tuesday’s deploy moved signups, what ChatGPT fetched this week.

Every tool returns a sentence and the numbers behind it, so an agent can quote the answer or reason over the table. Every answer is an aggregate over visits. There is no tool that returns a visit, a session or a person, and there never will be: the daily-rotating visitor id is a promise about what leaves the database.

What you need

An API key for the site with Read this site’s analytics checked. Add Write notes on charts too if you want the agent to be able to pin a note when it deploys something. The key names the site, so an agent connected with it sees that site and nothing else.

The endpoint is https://lynq.byharsh.com/api/mcp. It takes a bearer token, keeps no session, and answers each request on its own, so it works from a laptop, a CI job or a cloud agent alike. Connect an agent shows the configuration for each client; the tools lists what it can be asked.

What it is good for

• The question behind a number. “Why did bounce rate go up on Thursday?” is a notes call, a summary with compare, a breakdown by entry source, and a sentence. An agent does that in one turn.
• Checking a change. After a deploy, an agent with the notes scope pins “Deployed v2.4”, waits, and compares the days after with the days before.
• Writing it down. A weekly report, a changelog line, a Slack message: the agent reads the numbers and writes the prose, and the numbers are the dashboard’s.

Limits

A key may make 120 requests a minute. Tables are capped at 200 rows. Ranges are in the site’s timezone and go back as far as the site’s retention. A request carrying a browser Origin is refused, so a key that leaks into a web page cannot be used from it.