PostHog self-driving

What if your product

built itself?

You have a new pull request ready for review. (Yep, really.)

While you slept, PostHog dug through your product data, found what was worth fixing, and had agents do the work. All you need to do is hit merge.

PostHog instruments your codebase, then combines that context with product data like analytics events, errors, and recordings to understand problems and propose fixes.

How PostHog makes your product self-driving

Notable things happening in your product

A signal is a single observation about your product.

Support

Tickets and conversations from your users

PostHog agents run on their own, but don't run wild

Self-driving is autonomy from instruction, not from you. Agents work in the background without you prompting them to make progress, but nothing ships on autopilot.

• It's stuck in a sandbox. Work happens in the cloud, nowhere near your repo. Agents follow your branch protections, CI, and review rules.
• The work can't merge itself. Nothing reaches production until a human clicks merge. Robots don't touch the big green button.
• PRs > Issues. Raw signals are deduped and clustered into reports. Actionable ones become PRs. You only pay for real work completed.
• Your secrets are safe. Literally, and so is your code. Private repos stay private, and so does your data. (We're training AI models, but not on your code.)

See how it works

Works in your workflow

The same Inbox and agents show up wherever your team works.

PostHog DesktopBeta

A desktop app for driving parallel agents to edit your product. The same Inbox and reports live here.

PostHog Web

The full product in your browser. Explore your data, review proposed work, and dig into the evidence.

PostHog Slack

@PostHog brings reports, data answers, and agent work into the channels each team already watches.

PostHog MCP

Pull self-driving context into other tools, and pull context from your other tools into PostHog.

PostHog CLI

Self-driving from your terminal. Query your data and drive agents right where you already work, and wire it into your scripts and CI.

Give your agents the tools they need

PostHog gives your product the context it needs to become self-driving: usage, errors, replays, flags, experiments, feedback, logs, and the tools agents use to act on what they learn.

Everything lives in your context warehouse

Use PostHog as the full context layer for your product, or mix and match with your own tools.

Learn it by use case

The pocket guide to self-driving

A quick overview of self-driving, plus real use cases. Each one walks through the report a scout files, the pull request it becomes, and the scout itself. You can also add each use case as a custom scout to your own product.

Read the pocket guides

So, what's left for you?

Work lands while you sleep. You wake up to diffs and reports waiting for review. Then what?

Questions?

Answers

There are a few ways you can get answers to specific questions about self-driving products.

1. Check the docs

   Read how Inbox, signal sources, scouts, reports, and pull requests fit together.

2. Read a pocket guide

   Use case books that walk through real scouts – each one ends in a one-click way to add it to your product.

3. Ask PostHog AI

   Ask about the product, docs, examples, or how this fits your own workflow.

4. Talk to a human

   Useful for questions about setup, terms, volume, or whether self-driving fits your team.