Microsoft Clarity MCP server for your AI

With a single API token that you copy from Clarity. A project admin goes to Settings → Data Export → Generate new API token and pastes the value into the API token field on the connect form. There is no OAuth popup and nothing to re-authorize later, and the token is scoped to one Clarity project, so the connection is locked to that project.

The Data Export endpoint returns the last 1–3 days, and every tool on this connection inherits that window. It is built for what is happening on the site now — a source that went quiet, a page whose engagement fell, a control that started collecting rage clicks — rather than month-over-month trend work. Asking for last quarter will not return anything, because the API does not serve it.

No. Every tool on this integration reads; the Data Export API this connection uses has no endpoints that create, edit, or delete anything in your project. Read-only access mode is still the sensible default, and approval-required-for-writes simply has nothing to hold back here.

No. The tools return Clarity's Data Export insights — Traffic, Engagement Time, Scroll Depth, Popular Pages, and click-quality metrics such as rage clicks and dead clicks — as aggregated numbers broken down by dimensions like device, OS, country, source, and channel. Individual session replays and heatmaps are not part of the endpoint this connection reads; those stay in the Clarity dashboard.

Clarity's endpoint accepts at most three dimensions in a single request, so the tools are bundled to make each call count. clarity_insights_by_audience returns Device, OS, and Country/Region together, and clarity_insights_by_acquisition returns Source, Medium, and Channel together — one request each instead of three, which matters because the Data Export API is quota-limited. When you want a combination of your own, clarity_get_live_insights takes up to three dimensions you pick.

Threshold triggers cover sessions dropping, rage clicks spiking, dead clicks spiking, script errors spiking, engagement time dropping, bot traffic rising above a share of total traffic, a top page losing traffic, and traffic falling from a specific source, medium, or channel. You set the level for each one. Rage clicks and script errors are usually the first two worth wiring up, since both tend to mean a control on the page has broken.

No. PopMCP hosts it and gives you a Streamable HTTP MCP URL. There is nothing to deploy, no callback URL to register, and no infrastructure to keep running or patch.

Any client that supports a remote Streamable HTTP MCP server — Claude Desktop, Claude Web, ChatGPT custom connectors, Codex and the Codex CLI, OpenCode and Z.ai's ZCode, as well as editors such as Cursor, Windsurf and Zed. Your client signs in to PopMCP with OAuth; the provider credential never leaves the server.

Loading 5 tool definitions into every conversation burns context before the work starts. PopMCP loads a curated 5-tool set covering the everyday workflows and keeps the rest searchable, so your AI pulls a specialist tool only when a task needs it. You can switch a connection to the full catalog at any time.

A tool is something your AI calls — there are 5 here. A trigger is something PopMCP watches for you: 8 Microsoft Clarity events that can start work on the PopMCP side. Triggers never enter your AI client's context, and none of them ask you to paste a callback URL into Microsoft Clarity.