--- title: "PostHog MCP: review, setup & limitations · Undominated.ai" canonical: https://undominated.ai/mcp-servers/posthog/ description: "Query product data, investigate errors, and manage feature flags or experiments in a PostHog project." --- # PostHog MCP: review, setup & limitations · Undominated.ai > Query product data, investigate errors, and manage feature flags or experiments in a PostHog project. [← Explore all mcp servers](/mcp-servers/) PRODUCT ANALYTICS / PostHog # PostHog MCP Query product data, investigate errors, and manage feature flags or experiments in a PostHog project. See setup guidance ↓Original source ↗ SOURCE REVIEW Reviewed 2026-09-21 Evidence 6 linked sources Publisher PostHog Licence MIT (MCP service source; enterprise code excluded) ↗ Revision dac21711924e Read what was—and wasn’t—checked ↓ “Once connected, your agent can read and write across PostHog's products.” PostHog · upstream description ↗ Our analysis follows below. 01 / THE REASONING ## Why this made the selection. - Links analytics and error investigation with feature-flag and experiment workflows. - Documented project scoping, read-only mode and tool filters support narrower client access. ### A good fit for - Investigate a product error alongside usage data. - Review and adjust an explicitly approved feature flag or experiment. ### Weigh up before choosing - Write tools can change project behaviour, including flags and experiments. - Some tools use models internally and may incur PostHog AI spend; organization AI-data-processing settings affect availability. - Combining features and tools filters forms a union, so it can expose more tools than either filter alone. 02 / THE REVIEW RECORD ## What we actually inspected. Source review has boundaries. A clear record is more useful than a “safe” badge. ### Material inspected - PostHog MCP overview, FAQ/advanced setup and tool reference. - Pinned services/mcp README, package manifest, architecture and root licence. ### Our findings - OAuth routes the account to its US or EU region. A project-scoped API key is another documented option. - The readonly query parameter or x-posthog-read-only header excludes write tools; project headers or key scope are separate controls. ### Not established by this review - Server launch, authenticated tool calls, and release-to-source parity were not tested. The review applies to the material and revision named here. A newer upstream release can change its behavior. 03 / PUT IT TO WORK ## Connect a server deliberately. Upstream setup instructions ↗ DOCUMENTED COMMAND npx @posthog/wizard mcp add Copy command ↗ Copying does not execute this command. It may retrieve a newer version than the reviewed source. - Run the PostHog Wizard command below and select the intended client. - Authenticate to the intended PostHog account and project. - For an investigation-only session, use the documented readonly=true endpoint option; review flag or experiment writes separately. ### Before you start - PostHog account/project access. - Node.js/npx for the wizard, or manual remote-client setup. ### Compatibility Claude Code · Claude Desktop · Codex · Cursor · VS Code ### Implementation TypeScript ### Transports Streamable HTTP ### Remote OAuth or project API key - Reads authorized product analytics, errors, SQL results and project objects. - Can create, update or delete supported resources unless the connection is restricted to read-only tools. ### Cost model MCP access has no separate charge in the current guide; underlying service usage and AI-powered tools can incur account charges. 04 / FOLLOW THE EVIDENCE ## The source trail. Our notes are separate from the original resource. Check upstream before adopting a new version. - PostHog MCP setup and scope ↗ Checked 2026-09-21 https://posthog.com/docs/model-context-protocol Supports: summary, upstreamDescription, whySelected, bestFor, limitations, install, access, compatibility, transports, review - PostHog authorization, filtering and billing boundaries ↗ Checked 2026-09-21 https://posthog.com/docs/model-context-protocol/faq Supports: install, access, limitations, review - PostHog MCP tool inventory ↗ Checked 2026-09-21 https://posthog.com/docs/model-context-protocol/tools Supports: summary, whySelected, bestFor, access - PostHog MCP package manifest ↗ Checked 2026-09-21 https://raw.githubusercontent.com/PostHog/posthog/dac21711924ee4e8762c201906252eae0c8dbd4c/services/mcp/package.json Supports: language, review - PostHog MCP architecture ↗ Checked 2026-09-21 https://raw.githubusercontent.com/PostHog/posthog/dac21711924ee4e8762c201906252eae0c8dbd4c/services/mcp/ARCHITECTURE.md Supports: review, transports - PostHog licence scope ↗ Checked 2026-09-21 https://raw.githubusercontent.com/PostHog/posthog/dac21711924ee4e8762c201906252eae0c8dbd4c/LICENSE Supports: license KEEP COMPARING ## Other approaches to consider. Related by category or shared topics. These are alternatives to inspect, not a measured quality order. [### Google Analytics MCP ↗ A local connector for authorized Google Analytics property metadata and reporting APIs.](/mcp-servers/google-analytics/) [AI Tools ↗](/tools/)[Skills ↗](/skills/)[Agents ↗](/agents/)[MCP Servers ↗](/mcp-servers/)