[Analytics MCP servers](/about/analytics-mcp-servers) November 16, 2026

# Microsoft Clarity MCP server: setup, tools and limits

Microsoft’s Clarity MCP server runs locally with npx and signs in with your Clarity account. See its tools, what it can change and the limits to plan for.

![](/images/65d75b0ae8d69ef95785a5d6_david-pombar.webp)

**[David Pombar](/authors-guides/david-pombar)** Swiss army knife at Trackingplan

6 min read · 1327 words

> **TL;DR** To use the Microsoft Clarity MCP, add Microsoft’s open-source server, `@microsoft/clarity-mcp-server`, to Claude Desktop, VS Code or another MCP client; it runs on your computer through `npx`. Ask the client to sign in to Microsoft Clarity, finish the sign-in in the browser, pick a default project, and ask. It reads the dashboard, session recordings, AI Visibility data and Clarity’s docs. The old API token is being deprecated.

Clarity’s MCP server has changed faster than the pages that describe it. The version on GitHub signs in with your Clarity account and exposes 25 tools, from the dashboard to AI crawler activity. The page on Microsoft Learn, last updated in June 2026, still puts an API token on the command line. When they disagree, this guide follows the repository, which is the code you actually run.

This page is part of our series on [analytics MCP servers](/guides/analytics-mcp-servers). It covers what the server is, how to set it up, the tools it exposes, and the limits to plan around.

![Diagram in four steps: install Microsoft’s server with npx on your computer; sign in to Clarity in the browser and pick a default project; ask about the dashboard, session recordings, AI Visibility and Clarity’s docs; plan around the limits, which for the token-based setup are 10 requests per project per day, 3 days of data and 3 dimensions per request. A band says Clarity answers from what its tag recorded, so pages where the tag never loaded are missing from every answer.](/images/microsoft-clarity-mcp-steps.webp)

Local server, browser sign-in, one focused question at a time.

## What the Clarity MCP server is

It’s Microsoft’s own MCP server for Clarity, published as open source under the MIT license in [Microsoft’s clarity-mcp-server repository](https://github.com/microsoft/clarity-mcp-server) and documented on [Microsoft Learn’s MCP Server page](https://learn.microsoft.com/en-us/clarity/third-party-integrations/clarity-mcp-server). Microsoft describes it as an interface between Clarity analytics data and AI agents.

|     | Microsoft Clarity MCP server |
| --- | --- |
| Publisher | Microsoft, official |
| Package | `@microsoft/clarity-mcp-server` |
| Where it runs | On your computer, started by your client with `npx` |
| Sign-in | Your Clarity account, in the browser; Data Export API tokens being deprecated |
| Needs | Node.js 16 or later, admin or member access to at least one Clarity project |
| Clients named | Claude Desktop, Visual Studio Code, other MCP clients |
| Writes | Only AI Visibility Topic Insights reports and topics |

The repository documents no hosted version. Everything below assumes the server runs locally.

## How to set it up

The [repository’s README](https://github.com/microsoft/clarity-mcp-server#setup-and-installation) gives three routes.

**Claude Desktop, from the extension gallery.** Open ** File** > ** Settings** > ** Customize**, search ** Connectors** for ** Microsoft Clarity**, and select ** Install**.

**Claude Desktop or any client that takes a JSON config.** Add this to the client’s configuration. For Claude Desktop that file is `claude_desktop_config.json` in `~/Library/Application Support/Claude/` on macOS or `%AppData%\Claude\` on Windows. Then restart the client.

```json
{
  "mcpServers": {
    "microsoft-clarity": {
      "command": "npx",
      "args": ["-y", "@microsoft/clarity-mcp-server"]
    }
  }
}
```

**Visual Studio Code.** The README has an ** Install Server** button that adds the same `npx` command to VS Code.

Then sign in:

1.  Ask your client to **sign in to Microsoft Clarity**.
2.  The server opens Clarity’s sign-in page in your browser, or returns a link if it can’t.
3.  Finish the sign-in and go back to the client. It checks the status by itself.
4.  If you can access more than one project, pick a default one.

Microsoft’s own first prompts are good tests: “How many Clarity sessions did we get in the past 7 days?” and “List the most recent Clarity sessions from mobile devices.”

### Moving off the API token

Older setups pass `--clarity_api_token` in the command, which is still what Microsoft Learn shows. The README says token authentication “is being deprecated” and lists the migration: remove the argument and any `CLARITY_API_TOKEN` variable, restart the server, sign in, then delete the old token under **Settings** > ** Data Export** once the sign-in works.

## The tools it exposes

The README groups the tools by area:

| Area | Tools | What you get |
| --- | --- | --- |
| Dashboard | `query-analytics-dashboard` | Traffic, scroll depth, engagement time and more, filtered by browser, OS, country or device |
| Session recordings | `list-session-recordings` | Recordings filtered by URL, device, browser, OS, country, city and more |
| AI Visibility: citations | 6 tools, such as `get-citation-overview` and `list-top-cited-pages` | AI citations, AI referred traffic, queries, topics and cited pages for your verified domain |
| AI Visibility: bot activity | 9 tools, such as `list-ai-crawlers` and `get-scrape-to-referral-ratio` | AI crawler requests, paths, status codes and trends; needs a CDN connected in Clarity |
| AI Visibility: Topic Insights | 7 tools, from `list-topic-insights` to `delete-topic-insight` | Topic reports with citation rate and competitive share; needs a verified domain |
| Documentation | `query-documentation-resources` | Answers from Clarity’s own docs |

Only the Topic Insights tools change anything. `create-topic-insight-report`, `regenerate-topic-insight-report` and `delete-topic-insight` need the OAuth sign-in and enough permissions on the project, and a deletion has to be confirmed first. Everything else reads.

The README’s advice on how to ask is short: “keep each request focused on one thing.” A question that mixes page views, session length and conversions across segments is several questions.

## Limits to plan for

[Microsoft Learn’s MCP Server page](https://learn.microsoft.com/en-us/clarity/third-party-integrations/clarity-mcp-server) states the limits for the token-based setup: “up to 10 API requests per day, with a maximum of 3 days’ data and up to 3 dimensions per request”, per project. The [Data Export API reference](https://learn.microsoft.com/en-us/clarity/setup-and-installation/clarity-data-export-api) behind it adds that a response is limited to 1,000 rows and can’t be paginated. Microsoft lists higher API limits and multi-project support among the planned improvements.

The repository doesn’t publish limits for the newer sign-in. Until it does, plan as if the old ones held: one metric, one period and one breakdown per question.

Two prerequisites decide whether some tools return anything at all. Bot activity needs a CDN connected to the project, and Topic Insights needs the domain verified in Clarity. “Why am I not seeing any bot activity data?” is one of the README’s own example prompts.

## What Clarity can’t tell the model

Clarity answers from what its tag recorded. If the tag stopped loading on one template after a release, or a consent change cut it off in one market, those sessions are missing, and the model analyzes the rest as if it were the whole site. A drop in rage clicks can mean a better page or a page Clarity stopped seeing.

That is the part the Trackingplan MCP covers. Trackingplan watches the requests your site sends to [Microsoft Clarity](/integrations/microsoft-clarity) and the other analytics and ad tools on real traffic, the dataLayer underneath them, and the consent state, and flags dead tags, missing properties and consent breaches as they appear. Its agent answers from that evidence, and every answer shows the queries behind it.

Add the [Trackingplan MCP](/landings/mcp) to the same client and one chat can ask Clarity what visitors did and Trackingplan whether the tracking behind it held. Each person signs in with their own account, with no API key, and plan changes made from the chat wait for your approval. If you also run GA4, our guide to the [Google Analytics MCP server](/guides/google-analytics-mcp-server) covers Google’s side.

## FAQ

### How do I use the Microsoft Clarity MCP?

Add Microsoft’s server, @microsoft/clarity-mcp-server, to an MCP client such as Claude Desktop or VS Code; it runs on your computer with npx and needs Node.js 16 or later. Then ask the client to sign in to Microsoft Clarity, finish the sign-in in your browser, pick a default project and ask your question.

### Is the Microsoft Clarity MCP server official?

Yes. Microsoft publishes it as open source under the MIT license in the microsoft/clarity-mcp-server repository on GitHub, and Microsoft Learn documents it under Clarity’s integrations.

### Do I still need a Clarity API token for the MCP server?

Not with the current version. The repository says Data Export API token authentication for the MCP server is being deprecated and replaces it with signing in with your Clarity account. Microsoft Learn still shows the older token command, so follow the repository.

### Can the Clarity MCP server change anything in my project?

Almost everything it exposes reads. The exception is AI Visibility Topic Insights: it can create and regenerate topic reports and delete topics, which needs the OAuth sign-in and enough permissions on the project, and a deletion has to be confirmed first.

### What are the Clarity MCP server’s limits?

For the token-based setup, Microsoft Learn lists up to 10 API requests per project per day, data from the last 3 days at most, and up to 3 dimensions per request. The repository doesn’t publish limits for the newer sign-in, so ask small, single-purpose questions.

Part of[Analytics MCP servers: GA4, PostHog, Mixpanel and more](/guides/analytics-mcp-servers)

On this page

1.  [*01*What the Clarity MCP server is](#what-the-clarity-mcp-server-is)
2.  [*02*How to set it up](#how-to-set-it-up)
3.  [*03*The tools it exposes](#the-tools-it-exposes)
4.  [*04*Limits to plan for](#limits-to-plan-for)
5.  [*05*What Clarity can’t tell the model](#what-clarity-can-t-tell-the-model)
6.  [*06*FAQ](#faq)

**Stop babysitting tags**

Trackingplan audits your real user data around the clock and flags issues before they cost you.

[Start free trial →](https://app.trackingplan.com/signup)

![](/images/65d75b0ae8d69ef95785a5d6_david-pombar.webp)

**David Pombar** Swiss army knife at Trackingplan

Read more from David, a Senior Product Strategist with 18+ years in digital product development and an atypical error detection knack.

Share Copy link [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.trackingplan.com%2Fguides%2Fmicrosoft-clarity-mcp) [X](https://twitter.com/intent/tweet?url=https%3A%2F%2Fwww.trackingplan.com%2Fguides%2Fmicrosoft-clarity-mcp&text=Microsoft%20Clarity%20MCP%20server%3A%20setup%2C%20tools%20and%20limits)

Read next

[All Guides →](/guides)

-   [
    
    11 minAnalytics MCP serversNovember 16, 2026
    
    ### Analytics MCP servers: GA4, PostHog, Mixpanel and more
    
    The analytics MCP servers compared: GA4, PostHog, Mixpanel, Amplitude, Adobe, Clarity and browser servers. What each reads, changes and needs to connect.
    
    by **David Pombar**](/guides/analytics-mcp-servers)
-   [
    
    7 minAnalytics MCP serversNovember 16, 2026
    
    ### Google Analytics MCP server: tools, clients and limits
    
    What Google’s Analytics MCP server is, the 13 tools it exposes, the clients it supports, how it runs locally or hosted, and what it can’t see in GA4.
    
    by **David Pombar**](/guides/google-analytics-mcp-server)
-   [
    
    6 minGoogle Analytics 4October 26, 2026
    
    ### How to track ChatGPT and AI traffic in GA4
    
    Track visits from ChatGPT, Gemini, Perplexity and other AI assistants in GA4: the AI Assistant channel, what it misses, and the AI traffic that lands in Direct.
    
    by **David Pombar**](/guides/track-chatgpt-ai-traffic-ga4)

Trackingplan

## See everything. Miss nothing.

Your implementations audited around the clock with real-time, real-user data. Real-time alerts about errors or changes in your data, campaigns, pixels, privacy and consent. Let AI flag issues before they cost you.

[Start free trial →](https://app.trackingplan.com/signup) [Book a demo](/demo)
