PostHog – Open‑source product analytics & observability platform

What it does (in plain language)

PostHog is a self‑hostable service that collects everything a product team needs to understand how users interact with their software:

• Event analytics – record clicks, page views, purchases, etc., and explore them with charts or raw SQL.
• Web & mobile analytics – GA‑style dashboards, heatmaps, conversion funnels.
• Session replay – watch recordings of real user sessions, with an AI‑powered “Replay Vision” that can surface notable moments automatically.
• Observability – capture errors, logs, and distributed traces alongside the product events so you can see the full context of a problem.
• Feature management – feature flags and A/B‑style experiments that let you roll out changes to a subset of users and measure impact.
• User engagement tools – surveys, in‑app chat/support, and automated workflows triggered by product data.
• Data warehouse & pipelines – sync raw events to a managed warehouse, run transformations, and forward data to 25+ external tools.
• AI helpers – a built‑in “PostHog AI” that can answer natural‑language questions about your data, generate SQL, and even draft pull‑requests based on detected issues.

All of these pieces sit on top of a single data store, giving teams a unified “context layer” for product decisions.

Who might use it

• Product managers & growth teams that need fast, queryable analytics without relying on third‑party SaaS.
• Engineers who want to tie error tracking, logs, and traces directly to user actions.
• Customer‑success / support teams that benefit from session replays and AI‑summarised tickets.
• Start‑ups or privacy‑focused companies that prefer self‑hosting their analytics stack.

How it works (high‑level architecture)

1. Ingestion – SDKs (JS, React, Python, Go, etc.) or a simple HTTP API send events, errors, logs, and traces to a PostHog server.
2. Storage – Events are stored in a PostgreSQL database (or ClickHouse in larger deployments) and optionally replicated to a data warehouse.
3. Processing – Real‑time pipelines enrich events, compute aggregates, and feed downstream features like feature‑flag evaluation or AI analysis.
4. Query & UI – A web front‑end provides dashboards, SQL editor, heatmaps, session replay player, and admin tools. The same data is exposed via a REST/GraphQL API for custom integrations.
5. Extensibility – Plugins, data pipelines, and the Model‑Context‑Protocol (MCP) let external tools (including LLM‑powered agents) interact with the stored context.

Getting started

• Quick cloud trial – Sign up for a free PostHog Cloud account (no credit card) and get a generous free tier.
• Self‑hosted hobby deploy – One‑line Docker script (curl … | bash) sets up a minimal instance (≈4 GB RAM). Good for up to ~100 k events/month.
• SDK integration – Install the appropriate library (e.g., npm i posthog-js for web, pip install posthog for Python) or drop the JavaScript snippet into your site.
• Setup wizard – npx @posthog/wizard walks you through configuring ingestion, feature flags, and optional AI modules.

Notable strengths & limits

Strength	Limitation
Unified data model – analytics, errors, logs, and traces live together, making cross‑domain queries easy.	Self‑host scaling – hobby deployment is limited; larger volumes need ClickHouse, Kubernetes, or migration to PostHog Cloud.
Open source MIT license – you can run it entirely on‑premises.	Enterprise features – some advanced capabilities live in the ee directory (paid license) or the separate posthog‑foss repo.
AI integrations – Replay Vision, PostHog AI, and self‑driving reports automate insight generation.	AI is optional – requires OpenAI/Claude keys and may add latency/cost.
Rich plugin ecosystem – data pipelines, third‑party integrations, and custom endpoints.	Learning curve – full product‑OS includes many modules; teams may start with core analytics and add pieces over time.

Bottom line

PostHog is a full‑stack, open‑source platform that turns raw product events into actionable intelligence, blending traditional analytics with modern observability and AI‑driven insights. It’s especially appealing for teams that want control over their data, a single source of truth for product context, and the flexibility to extend the system with custom pipelines or LLM agents.