PostHog is an open-source platform for building self-driving products, combining product analytics, web analytics, session replay, feature flags, experiments, error tracking, logs, surveys, a data warehouse, data pipelines, AI observability, and workflows in a single system, aimed at product, engineering, and growth teams that want to own their product data rather than rent it.

What it is

PostHog is an open-source analytics and product-tooling platform written primarily in Python, with JavaScript and React components for client-side instrumentation. It lives in the web and product analytics ecosystem, and it also reaches into session replay, experimentation, error tracking, log ingestion, customer data pipelines, and LLM observability. Rather than stitching together a separate tool for each of those jobs, PostHog captures event data once and lets teams query it alongside data synced from external sources such as Stripe and Hubspot.

The concrete problem it solves is fragmentation. Teams typically run one paid service for event analytics, another for session replay, another for feature flags and experiments, and another for error tracking, none of which share a data model. PostHog replaces that stack with a single self-hostable or cloud-hosted platform where analytics, replays, flags, experiments, errors, logs, and surveys read from the same event stream. The README explicitly positions it as the open alternative to Mixpanel, Amplitude, and FullStory.

Key capabilities

• Product analytics with autocapture or manual event instrumentation, plus analysis through visualizations or SQL.
• Web analytics dashboard in the style of Google Analytics, covering conversion, web vitals, and revenue.
• Session replays of real user sessions for websites and mobile apps, used to diagnose issues and understand behavior.
• Feature flags for rolling out features to selected users or cohorts, and experiments that measure statistical impact on goal metrics with a no-code setup option.
• Error tracking with alerts and issue resolution, plus log ingestion, search, and analysis next to the rest of the product data.
• Data warehouse sync from external tools including Stripe and Hubspot, queryable alongside product data, and data pipelines (CDP) that filter and transform incoming data, send it to 25+ tools or any webhook, or batch export to a warehouse.
• AI observability for LLM-powered apps, capturing traces, generations, latency, and cost, with control available from Slack, web, PostHog Desktop, or an editor via the MCP.

Who uses it and how

• Product and growth teams that need event analytics, funnels, and experiments against one shared dataset instead of reconciling exports from separate vendors.
• Engineering teams using feature flags and error tracking to release changes to cohorts gradually and catch regressions in production.
• Support and design teams watching session replays to reproduce reported bugs and understand where users get stuck.
• Data teams syncing Stripe, Hubspot, or warehouse data into PostHog so product events and business metrics can be queried together.
• Smaller teams starting on PostHog Cloud, where the first 1 million events, 5k recordings, 1M flag requests, 100k exceptions, and 1500 survey responses are free every month.

Getting started

The README recommends signing up for PostHog Cloud US or PostHog Cloud EU as the fastest and most reliable route. For self-hosting, it documents an advanced one-line Docker hobby deploy on Linux, with 4GB of memory recommended.

How it compares

PostHog is presented as the open-source alternative to Mixpanel, Amplitude, and FullStory, which are closed hosted products. PostHog offers both a hosted option and a self-hosted Docker hobby deploy, so teams can keep analytics, replay, and flag data inside their own infrastructure rather than sending it to a vendor. On cost, the free tier applies to every tool with usage-based pricing beyond it, whereas the products it replaces are subscription services with their own per-event or per-seat pricing.

When to use it — and when not to

Self-hosters take on operating a Docker deployment with roughly 4GB of memory, and the README itself labels the open-source hobby deploy advanced and recommends cloud instead, so a team without container operations experience should start hosted. The repository metadata carries a NOASSERTION licence value, which means the exact licence terms are not machine-readable here and should be checked before any commercial redistribution. Teams wanting a small, single-purpose analytics library rather than a broad platform, or those unwilling to accept a fast-moving project with thousands of open issues, should look elsewhere.

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/posthog/posthog/HEAD/bin/deploy-hobby)"