PostHog MCP Server: Give AI Agents Access to Analytics, Feature Flags, and Session Replay

Why Give AI Access to PostHog?

PostHog combines product analytics, feature flags, session replay, and A/B testing in one platform. When building features, you tab over to PostHog to check conversion funnels, verify flag rollouts, or watch session replays. Each switch costs context.

An MCP server for PostHog lets your AI agent query analytics events, check flag configurations, pull funnel data, and surface replay insights directly from Claude, Cursor, or Windsurf through DataFaucet.

What Becomes Possible

Once your AI agent has PostHog access:

• "What's the conversion rate on the signup funnel this week vs last?"
• "Is the new-checkout-flow feature flag enabled for the US cohort?"
• "Which events are users firing before they churn?"
• "Show me the top 5 paths users take after landing on /pricing"
• "What's the p95 page load time from web vitals events?"
• "Are there session replays with rage clicks on the settings page?"

How It Works

1. Go to DataFaucet and enter your PostHog instance URL (app.posthog.com or self-hosted)
2. Authenticate with your PostHog credentials
3. Browse dashboards, funnels, flags, and replays as you normally would
4. Each API call becomes an MCP tool your AI can invoke
5. Deploy and connect to Claude, Cursor, or Windsurf

The AI agent accesses whatever project data your PostHog permissions allow.

Security Best Practices

• Read-only API key — use a personal API key scoped to read operations only
• Project isolation — limit access to a single PostHog project, not all organization projects
• No flag writes — prevent AI from enabling or disabling feature flags in production
• No data deletion — block AI from deleting events, persons, or cohorts
• No billing access — exclude organization settings and billing endpoints
• Audit via activity log — PostHog logs all API access for compliance review

Use Cases by Role

Product engineers: Feature context. "What percentage of users saw the new onboarding? What's their retention vs control?" while deciding whether to ship or iterate.

Growth engineers: Funnel debugging. "Where are users dropping off in the upgrade flow? Show me the step-by-step conversion rates" before optimizing the code.

Frontend developers: Performance context. "What's the LCP distribution for /dashboard? Are web vitals regressing after the last deploy?" while working on rendering.

Product managers: Decision support. "Compare activation rates between cohorts who used feature X vs those who didn't. What events correlate with 7-day retention?"

PostHog UI vs AI Agent

Product Stack Integration

PostHog connects with your deploy pipeline and experimentation workflow. With DataFaucet MCP servers for each:

• Correlate feature flag rollouts with error rate changes in Sentry
• Check PostHog funnel impact after deploying via Vercel or GitHub Actions
• Cross-reference analytics events with Datadog latency metrics

Getting Started

Set up read access for your AI agent:

1. In PostHog, create a personal API key with read-only permissions:

- Insights: Can query trends, funnels, paths, and retention

- Feature Flags: Can read flag configurations and rollout status

- Session Replay: Can list and filter recordings

- No create, update, or delete permissions

1. Point DataFaucet at your PostHog instance
2. Browse dashboards, funnels, and flags as you normally would
3. Deploy and connect to your AI coding tool

Related: Datadog MCP Server for APM, Sentry MCP Server for error tracking, Best MCP Servers for the full directory.