Mixpanel MCP Server

The Mixpanel MCP Server connects AI assistants such as Claude, Cursor, or the Gemini CLI to the product analytics platform Mixpanel. Anyone who uses Mixpanel to measure how people use an app or website can ask the assistant questions like "How has checkout conversion developed over the last four weeks?" or "Which user groups drop off after onboarding?" without clicking queries together in the web interface. The server is operated by Mixpanel itself. According to the documentation it runs hosted as a remote server, and individual features are marked as beta.

What is the Mixpanel MCP Server?

Mixpanel stores events and user profiles and exposes them through its own analyses. The MCP server makes these functions accessible to assistants. According to the documentation the address is regional: https://mcp.mixpanel.com/mcp for the US region, https://mcp-eu.mixpanel.com/mcp for the EU, and https://mcp-in.mixpanel.com/mcp for India. You have to choose the address that matches the data center of your project, otherwise the assistant will not find your data.

Which tools are available?

The documentation names roughly 60 tools in about twelve groups. They include queries and reports on funnels, flows, and retention, access to dashboards, exploration of existing events and properties, and maintenance of the Lexicon, the data catalog with descriptions. Cohorts, lookup tables, metrics, experiments, and feature flags are added; audit log access and some further areas are marked as beta. According to the documentation, session replays can also be included.

What is it particularly useful for?

The value lies in the quick path from a question to a number. Product teams that used to wait for analysts can check hypotheses directly, and developers see in the editor how a release affects usage data. Compared with a generic database server, this server knows the platform's concepts, such as events, cohorts, and experiments, and answers within the scope of Mixpanel permissions. Tidying up the Lexicon with the assistant also saves time.

Setup

Sign-in works via OAuth in the browser with your own Mixpanel account. In Claude you add the Mixpanel connector or enter the matching regional address under custom connectors. In Cursor, according to the documentation, the proxy mcp-remote with the address is registered in the mcp.json file, which requires Node.js. For automation and CI/CD there are service accounts in beta with a static authorization header; the credentials belong in a secret manager and not in a configuration file in the repository.

Requirements, security, and limits

An organization administrator has to enable MCP in the organization settings; for new Free and Growth accounts it is on by default since August 1, 2026, according to the documentation. Users only see projects they already have rights to. The assistant can not only read but also write, for example in the Lexicon; changes land in the audit log. Retrieved data enters the AI client context and therefore reaches the model provider. Named limits are at most 600 requests per hour and user as well as missing cohort filters in the Run-Query call.

Costs

The server requires a Mixpanel account. According to its pricing page the platform offers a permanently free plan, a paid Growth plan, and an Enterprise tier; query limits may apply to the free plan. Which terms apply to you is shown on the official Mixpanel pricing page.

Who is the Mixpanel MCP Server for?

It is worthwhile for product, growth, and engineering teams that already use Mixpanel and want to do analyses in conversation with an assistant instead of in the web interface. Anyone without Mixpanel data gains nothing from it: the server is not a standalone analytics tool but an access path to the platform. Because of the write permissions, teams should also clarify who may use the assistant.

A Mixpanel account with access to at least one project, MCP access enabled by an administrator, and an MCP client that supports remote servers, such as Claude or Cursor.

With the Mixpanel MCP Server, an AI assistant analyzes your product data. This guide leads from activation to the first query.

Step 1: Check the prerequisites

You need a Mixpanel account with access to at least one project and an MCP client that supports remote servers, such as Claude or Cursor. An organization administrator has to enable MCP in the organization settings; for new Free and Growth accounts it is on by default since August 1, 2026, according to the documentation.

Choose the region

The address depends on the data center of your project: https://mcp.mixpanel.com/mcp for the US, https://mcp-eu.mixpanel.com/mcp for the EU, and https://mcp-in.mixpanel.com/mcp for India. With the wrong region the assistant does not find your projects.

Step 2: Set up in Claude

Add the Mixpanel connector via the Claude directory and complete the OAuth sign-in in the browser. For the EU or India, enter the matching address under custom connectors.

Step 3: Set up in Cursor

In the file ~/.cursor/mcp.json, register a server with the command npx and the arguments -y, mcp-remote, and the address. On first connection Cursor asks for authorization at Mixpanel. Node.js must be installed.

Automation with service accounts

For CI/CD there are service accounts in beta with a static authorization header made of Base64-encoded credentials. Store them in an environment variable or a secret manager, never in the repository.

Step 4: First queries

Start with simple questions: "Which projects can I see?", "Which events exist?", or "Show the funnel from signup to purchase for the last 30 days." Spot-check results against the web interface.

Permissions and limits

The assistant only sees your projects but can also write, for example in the Lexicon; changes appear in the audit log. At most 600 requests per hour and user apply. Retrieved data reaches the model provider.

claude mcp add --transport http mixpanel https://mcp.mixpanel.com/mcp

OAuth via the browser with your own Mixpanel account; for automation, service accounts (beta) with a static authorization header.

Read and write access to the projects the signed-in user is authorized for; changes are recorded in the audit log.

The server runs hosted by Mixpanel in the chosen region. Retrieved data enters the AI client context and therefore reaches the model provider.

Unwanted changes in the Lexicon through write permissions, passing usage data to the model provider, service account credentials stored in the repository, and choosing the wrong region.