PostHog is an open-source product platform that combines analytics, session replay, feature flags, experiments, error tracking, logs, surveys, a data warehouse, and AI observability in one place. Because all of these share one data store, it positions itself as the context source agents need to diagnose problems and ship fixes in a product.
Its self-driving mode turns signals in product data, such as errors, rage clicks, and failed queries, into researched reports and pull requests that a person reviews and merges. The platform can be steered from Slack, the web app, the PostHog Desktop app, or an editor through its MCP server, which brings PostHog into Claude Code, Cursor, or any MCP-compatible agent.
Data comes in through a JavaScript web snippet, an API, or SDKs for frontend (JavaScript, Next.js, React, Vue), mobile (React Native, Android, iOS, Flutter), and backend (Python, Node, PHP, Ruby, Go, .NET) code. PostHog runs as a hosted service in US and EU regions with a monthly free tier per tool, or as a self-hosted hobby deployment on Docker sized for roughly 100k events a month. A separate fully FOSS edition strips out the proprietary code.
Features
- Self-driving mode: turns errors, rage clicks, and failed queries into researched reports and pull requests for review
- Product analytics: autocaptured or manually instrumented events, analyzed with visualizations or SQL
- Web analytics: traffic, sessions, conversion, web vitals, and revenue on a dashboard
- Session replay: recordings of real user sessions on websites and mobile apps
- Feature flags and experiments: staged rollouts to users or cohorts, and experiments measured against goal metrics
- Error tracking and logs: error alerts and log search alongside product data
- Surveys: no-code templates or a custom survey builder
- Data warehouse: syncs data from tools such as Stripe and HubSpot for querying with product data
- Data pipelines: filters and transforms incoming events and sends them to 25+ tools, webhooks, or a warehouse
- AI observability: traces, generations, latency, and cost for LLM-powered applications
- Workflows: automated actions and messages to users
- MCP server: query and operate PostHog from Claude Code, Cursor, or other MCP-compatible agents
