Woodshed

Your agent changed 1,284 lines. 3 decisions need you.

Woodshed tests every PR in a running copy of your app, has Claude and Codex review it, and hands you the decisions that need a person, in plain language.

curl -fsSL https://woodshed.run/cli | sh

free credits to start · open source · runs on your Claude Max or ChatGPT Pro plan

How 1,284 lines become 3 decisions

9 PRs, 1,284 lines written by Claude Code and Codexin
Booted in a real copy of your app

Each PR on its own machine, with your full environment: services, database, seed data.

Clicked through by a QA agent

The way a user would, not just the unit tests.

Reviewed by Claude and by Codex

Two models, two sets of blind spots.

Settled what they agree on

6 PRs come back ready to merge with nothing to ask.

3 plain-language decisions for youout
01

Tested, not just read

A prompt can review a diff. It can't boot your stack and click through the feature. Woodshed does both, for every PR.

02

Two reviewers, one verdict

Claude and Codex review each change. Where they disagree is where you look.

03

Questions, not diffs

What needs a person comes back as plain questions about behavior, with the evidence one click away.

04

On the plan you already have

Reviews and QA run on your Claude Max or ChatGPT Pro subscription. No API keys.

Each PR runs on its own isolated machine. Secrets are never logged. Woodshed only sees the repos you choose.

Stop reading. Start directing.

Start free with GitHub

We ship our own product with Woodshed every day.

or self-host it: paste this into Claude Code or Codex

Set up https://github.com/ChifferCo/woodshed for me

Or set it up by hand.