// the take
PostHog is a self-hostable product analytics platform that bundles feature flags, session replay, A/B testing, error tracking, surveys, and a data warehouse into one stack. It targets product-focused engineering teams who want to avoid stitching together Mixpanel, LaunchDarkly, Sentry, and a CDP. The open-source core is genuinely usable, though the interesting features mostly live in the paid tier.
The event pipeline architecture is well thought out — ClickHouse for analytics queries means you're not fighting Postgres at scale the way most self-hosted tools do. The HogQL layer (PostHog's SQL dialect) lets you drop into raw queries without leaving the UI, which is actually useful when the pre-built funnels don't fit your data model. The `.agents/skills/` directory shows they're investing seriously in AI-assisted development workflows, not just bolting on an LLM chatbot. SDK coverage is broad and consistently maintained — the React Native and Flutter SDKs aren't afterthoughts.
Self-hosting caps out around 100k events/month before they recommend migrating to their cloud, which makes the self-hosted path feel like a demo rather than a real deployment option. The `ee/` directory with its own license means you're not getting the full feature set under MIT — feature flags multivariate, group analytics, and cohort-based experimentation all require the paid plan. The monorepo is enormous (Python Django backend, TypeScript/React frontend, Rust ingestion services, Go services, Hog interpreter) and local dev setup is genuinely painful; the skill docs acknowledge this with a dedicated `run-posthog` skill file. Data warehouse connectors are useful but the sync reliability for self-hosted instances isn't well-documented.
We read the README, the directory tree and a key source file in June 2026. Nobody pays to be listed. How we judge →