PostHog Freemium

Product analytics, session replay and AI observability in one open-source stack

What is PostHog

PostHog is an open-source product analytics platform that bundles event tracking, session replay, feature flags, experiments and LLM analytics — replacing four separate vendor contracts with one SDK.

James Hawkins and Tim Glaser started it in 2020 after being rejected from Y Combinator’s summer batch; they got in the next time round and went on to open-source the whole stack. You can self-host on a single Docker Compose file, though the company will tell you the managed cloud is the sane choice above modest traffic.

The pitch is consolidation. Amplitude for funnels, Hotjar for replay, LaunchDarkly for flags, and some LLM tracer — that is four invoices and four SDKs leaking into your bundle.

Product Features

  • Product analytics — events, funnels, retention curves, path analysis and cohorts, queryable through a SQL interface.
  • Session replay — records browser sessions with console logs and network requests; 5,000 free recordings per month.
  • Feature flags — percentage rollouts, A/B experiments and remote config from the same dashboard.
  • Free tier — 1 million events, 5,000 replays and 100 flags monthly, no credit card.
  • Pricing — usage tiers; product analytics beyond the free allowance runs about $0.00005 per event at low volume and drops with scale.
  • LLM analytics — traces, latency, cost per call and prompt capture for OpenAI, Anthropic and LangChain integrations.
  • Self-hosting — MIT-licensed core; requires ClickHouse, Postgres, Redis and Kafka, so plan RAM accordingly.
  • Autocapture — records clicks and form submissions without manual instrumentation.
  • Data appetite — replay plus autocapture means real PII risk; masking rules need configuring before launch, not after.

Application Scenarios

  • Onboarding funnels — see exactly which signup step loses 40% of users, then replay ten sessions from that step.
  • Feature launches — flag-gated rollout to 10% with an experiment on conversion rate.
  • LLM cost control — trace token spend per feature and per customer without adding a separate observability vendor.
  • Support triage — jump from a user ID to their recent session and events.
  • Privacy-first deployments — self-host in your own cloud when a DPA with a US vendor is a non-starter.
  • Where it does not fit — pure marketing attribution across ad channels, where tools like GA4 or Amplitude’s marketing suite still fit better. Self-hosting is genuinely heavy; do not pick it if you have no one to run ClickHouse.

Quick Summary

  • PostHog bundles analytics, session replay, feature flags and LLM tracing in one platform.
  • The free tier includes 1M events, 5,000 replays and 100 flags per month.
  • It is MIT-licensed and self-hostable, though it needs ClickHouse, Kafka and Redis.
  • Best for product teams consolidating vendors; not a marketing attribution tool.