baum: human spaces for agentic teams

Review features.decisions.risks.features.
Not diffs.

baum brings your agents’ code, sessions and decisions into one place, and shows which parts need you.

Download baumFor macOS and LinuxDownload for macOSmacOS 12+, Apple silicon and IntelDownload for Linux (AppImage)x64 (Intel and AMD)Download for Linux (AppImage)arm64, for this computer’s ARM processorDownload for LinuxChoose the build for your computerNot available on Windowsbaum runs on macOS and LinuxNot available on iOSbaum runs on macOS and LinuxNot available on iPadOSbaum runs on macOS and LinuxNot available on Androidbaum runs on macOS and LinuxNot available on ChromeOSbaum runs on macOS and LinuxMore platforms

Fully open source. View the code on GitHub.

review used to be the slow part

Don’t be the bottleneck.

Your agents ship faster than anyone can review. Reading every diff and chasing every decision is what slows the team down. baum sends you only the calls that need your judgement, so your review keeps up with your agents.

what is baum?

See what your agents built, and what needs you

baum sits on top of your coding agents and review tools. It pulls the decisions out of every session, so your team can see how each feature is being built and who needs to look at what.

Read the docs →
  • Feature review — PRs grouped into features, with the decisions behind them.
  • Decision log — The reasoning from the agent session behind any piece of code.
  • Test coverage — How well each change was tested, before you look.
  • Triage — Only the work that needs human judgement comes to you.
  • Open source — Runs on your own computer, and the code is on GitHub for anyone to read.
  • Sessions that need you — See which agent sessions are waiting on you, and jump straight to their terminal or Claude app in one click.
  • Any model — Claude Code and Codex sessions, whatever the model.
  • Alongside your review agents — They check the code works. baum checks it’s the right thing.
  • GitHub-native — Connect your GitHub account. baum only writes to GitHub when you press a button.
questions

Before you ask.

Didn’t find what you were looking for? Write to us at hi@usebaum.com, we’re happy to help.

What is baum?

A review space that sits on top of your coding agents. It pulls the decisions out of every session and groups PRs into features, so your team sees how each one is being built and only reviews what needs a human.

Is baum another review agent?

No. Review agents check that the code works. baum shows what was decided and why, so you can check it’s the right thing. It runs alongside the review agents you already use.

Which agents does baum capture?

For now, Claude Code and Codex, whatever the model.

What data does baum use?

The agent sessions in the repositories you connect (prompts, agent replies and tool calls) and the pull requests and commits that come out of them.

Where is my data stored?

On your computer. The open-source app has no account and no server, so nothing it stores is ever sent to us. Reviews run through your own Claude Code or Codex, as when you use them yourself.

Who can see a session?

Only you: your sessions stay on your computer.

How much of my plan does baum use?

Very little. Each review is one short run of your own agent, capped at 10 minutes, and the app shows what every review used. You can pause automatic reviews at any time.

What does baum cost?

The open-source app is free under the MIT licence. Reviews use your own Claude Code or Codex plan.

Want to use baum with your team? Send us a message.

tap to play · hold to jump higher