Skip to directory

About PostHog MCP

A hosted MCP server for using PostHog analytics, SQL, error tracking, feature flags, experiments, and other enabled products from an AI client.

What you can use it for

  • Query product analytics and HogQL
  • Investigate errors and session replays
  • Manage feature flags and other configured PostHog workflows
Connection and setup
Hosting
hosted
Transport
Streamable HTTP
Endpoint
https://mcp.posthog.com/mcp
Authentication
PostHog OAuth hosted auth routes the user to the correct US/EU region.

Requirements

  • A PostHog account with access to the relevant project and product features.
  • An MCP client that supports remote HTTP/Streamable HTTP servers and OAuth.

Setup steps

  1. Use the official wizard command `npx @posthog/wizard mcp add` and choose the target client; or manually add the remote endpoint https://mcp.posthog.com/mcp using that client�s documented HTTP MCP configuration.
  2. When the client prompts, sign in to PostHog and authorize access for the selected account.
  3. Reconnect and confirm that tools are listed. Available tools and data depend on account, project, permissions, and enabled features.
  4. For AI-backed tools, confirm the organization permits AI data processing and review any AI spend implications before use.
Tools 3+

Three exact example tool names are shown in PostHog’s official MCP server reference. The default server mode may expose a single exec tool; full per-tool exposure and allowed names depend on mode, feature flags and client configuration.

  • dashboard-get

    Example stable PostHog dashboard tool; full tools mode and availability depend on client configuration.

    01
  • feature-flag-get-all

    Example stable PostHog feature-flag tool.

    02
  • execute-sql

    Example stable PostHog SQL query tool.

    03

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

Limitations and notes
  • Some AI-backed MCP tools may consume PostHog AI spend and require AI data processing to be enabled.
Original sources

Documentation checked 2026-10-28T19:00:00.000Z. Connection has not been runtime tested.

Similar tools