Give your agent
a second opinion.

The CLI that independently checks a proposed change, then returns the evidence your agent needs to continue.

Install Critique ↘Read the protocol →

C Critiqueyour agent's second set of eyes

Choose judgment model3 available

Claude Fable 5Anthropic

GPT-5.6-SolOpenAI

Kimi-K3Moonshot AI

Download CLI film ↓

OpenCodeCodexClaude CodeCursor

ChangeJudgmentEvidence

EARLY ACCESS · GPT-5.6-TERRA

A merge-minded CLI for teams building with agents first.

Hundreds of millions of GPT-5.6-Terra-powered tokens are ready for early users. Bring your coding agent; Critique brings the independent finish.

Claim early access ↗

Your agent writes the change.
Critique challenges the claim.

Different context. Different incentives. A result your agent can act on—not another thread for someone to interpret.

THE HANDOFF

Your agent keeps the wheel.
Critique checks the claim.

Critique does not replace your coding agent. It provides an independent pass and returns structured evidence for the next decision.

ONE AGENT. TWO CONTEXTS. ONE CLEAR DECISION.live collaboration trace

Your agentClaude Code · GPT-5.6-Sol · Cursor

Assigned taskStop duplicate charges after retry

✓candidate change written

CCritiqueindependent context · fresh verifier

01Inspectreconstructs the candidate change

02Evidencetests claims in fresh context

03Decisionreturns a stable judgment

RETURN TO CALLEREvidence packageresult · reproduction · 18 passing tests · proof

↗

✓Your agent receives an independent result.It decides the next move with evidence it did not generate itself.READY TO CONTINUE

Download collaboration film ↓

One call.
A second set of eyes.

Critique returns a judgment, its evidence, and a stable handoff your agent can use without scraping a dashboard.

Three connected Critique modules finding a defect, preparing a patch, and verifying it

Inspect

Reconstructs the change independently and tests the claims behind it.

Decide

Returns a clear result, the evidence behind it, and the limits of that result.

Continue

Hands structured proof back to the agent that owns the next move.

$ critique finish --repair none --jsonstructured result ready

Built for the agent making the call.

No dashboard ritual. Your coding agent calls one command, consumes structured output, and decides what happens next.

See the CLI contract →

agent callfin_7f3a

$ critique finish --repair none --json

→ inspection   change reconstructed
→ evidence     tests and limits attached
→ decision     safe to continue

✓ VERIFIED

fresh contextproof attachedexit 0

Three commands.
One independent check.

Install it once, add it to the agents you use, then call Critique whenever a change needs an independent decision.

Authorize once

critique login

Scoped machine credential

HUMAN STEP

Connect agents

critique integrate --agent all

Codex, Claude Code + OpenCode

ONE COMMAND

Check a change

critique finish --repair none --json

Structured judgment + evidence

AGENT LOOP

USE THE COMPUTE YOU ALREADY HAVE

ChatGPT in.
Critique out.

Authenticate through the official Codex CLI. Your OpenAI credentials stay there. Critique gets the agent, not your password.

critique setup --with-chatgpt

INSTALL THE CLI

Put an independent check in every agent loop.

$npm install --global @critiquedotsh/cli

v0.1.1Node 20+macOS · Linux · Windows

Then authorize one device and add Critique to the agents you already use.

critique login
critique integrate --agent all

Download package ↓View on npm ↗

Give your agenta second opinion.