Official remote MCP server hosted by Mixpanel
What it does
Mixpanel hosts a Model Context Protocol server that gives AI assistants direct access to your Mixpanel data. Users can query analytics using natural language without dashboard navigation — insights, funnels, flows, and retention, plus dashboard management, data discovery, and session replay analysis.
How to connect
Choose the endpoint for your deployment region:
| Region | Endpoint |
|---|---|
| US (default) | https://mcp.mixpanel.com/mcp |
| EU | https://mcp-eu.mixpanel.com/mcp |
| IN | https://mcp-in.mixpanel.com/mcp |
claude mcp add --transport http mixpanel https://mcp.mixpanel.com/mcp
After adding, run /mcp to get the OAuth link and sign in with Mixpanel. Auth uses the Authorization Code + PKCE flow — users authenticate with their own Mixpanel credentials, so existing project permissions and roles apply.
Prerequisite
An organization admin must enable MCP in Settings → Org → Overview before anyone can connect.
Capabilities
- Analytics — run insights, funnels, flows, and retention queries; retrieve saved reports; access query schemas
- Dashboards — create, list, retrieve, update, duplicate, and delete dashboards
- Data Discovery — browse projects, events, properties, property values, event metadata, and data quality issues
- Data Management — edit event/property metadata, create/rename/delete tags, dismiss data quality issues
- Session Replays — analyze user replay data alongside event history
Use cases
- “Compare the signup funnel conversion week-over-week for the last 30 days” — runs a funnel query and summarizes the insight
- “Did any activation events drop after this release?” — combines insights with event metadata to flag anomalies
- Create or update a Mixpanel dashboard as part of merging a PR
Notes
- Rate limit: 600 MCP requests/hour per user
- HIPAA is not currently supported — do not use for sensitive data in regulated environments
- Query results are sent to the AI provider you’re using — review against your data governance policies
- OAuth is required, so fully headless / automated environments aren’t supported
- Pick the endpoint that matches your organization’s deployment region — using the wrong one will look like empty data