Critique Code
A model-flexible coding harness that writes and tests code in the implementation loop.
Critique.shBetter tools for building software.
An agentic coding harness and an independent review CLI. Build the change, examine the evidence and keep the next decision in your hands.
Visit critique.shA distinct point of view.
Critique Code is the authoring harness: it implements, reviews and repairs code in one workflow. The separate Critique CLI checks local changes for an existing agent and returns structured evidence, limits and a handoff. Verification belongs inside Critique.
Software engineers and teams working with AI coding agents.
From a question
to something useful.
Start with
A software change that needs an independent check.
Work through
Reconstruct the change, examine the intent and run useful checks.
Leave with
Findings, evidence and a clear handoff to the person or agent deciding what happens next.
What it does.
Independent review
A fresh review context examines changes and checks the intent behind them.
Local-change CLI
The critique command works with staged, unstaged, deleted and untracked changes.
Evidence and handoff
Structured outcomes, reproduction evidence, explicit limits and repair packages when warranted.
Inside Critique.sh.
Original product capture. Features and availability are maintained on the official product site.