Critique/docs
Platform

@critique Bot

How to talk to the Critique bot in GitHub PR comments — supported commands, mention syntax, and configuration.

Critique is not just an automated gate — it's a conversational participant in your pull requests. Once the GitHub App is installed, you can @mention the bot in any PR comment to ask questions, request explanations, or trigger analysis on demand.

Mentioning the bot

Type @critique anywhere in a PR comment or inline review comment. The bot reads your full PR diff, relevant repository files, the PR description, and the comment thread before responding.

@critique why did we switch from useEffect to useSyncExternalStore here?

The bot responds directly in the thread, typically within 3–6 seconds.

Where it works

@critique mentions work in both general PR comments and inline review comments attached to specific lines. For inline comments, Critique automatically includes the file path, diff hunk, and line number in its context.

Supported commands

SyntaxWhat it does
@critique <question>Ask anything about the PR, the code, or the diff
@critique /reviewRe-run the full multi-agent review pipeline (Scout → Specialists → Lead)
@critique /ask <question>Explicit ask — same as a plain mention
@critique /explainExplain the changes in the PR, or explain the code at the commented line
@critique /fixSuggest a concrete fix for the issue being discussed
@critique /improveSuggest architecture or quality improvements
@critique /securityRun a security-focused analysis on the diff
@critique /testsSuggest missing tests for the changed code
@critique /helpShow the command reference in the PR thread

How GitHub App bot mentions work

GitHub App bots are registered accounts in the GitHub system. When your Critique GitHub App is installed, GitHub creates a bot account whose username matches the App's slug. To tag it, use @<slug> (e.g. @critique or @critique-bot).

Important: GitHub App bot accounts appear in comments with a [bot] badge, but you do not include [bot] in your @mention. Write @critique, not @critique[bot].

The bot listens on two webhook event types:

  • issue_comment — general PR-level comments in the conversation thread
  • pull_request_review_comment — inline review comments attached to specific diff lines

Both are handled automatically with no configuration required.

Configuration

Bot name: By default, Critique matches @critique and @critique-bot. If you self-host the app under a different slug, set the GITHUB_APP_BOT_NAME environment variable to your app's slug:

GITHUB_APP_BOT_NAME="my-critique-app"

The bot will then respond to @my-critique-app mentions.

Chat model: The bot uses Qwen 3.6 Plus (qwen/qwen3.6-plus) by default for its 1 million token context window. To override:

CRITIQUE_PR_COMMENT_CHAT_MODEL="openai/gpt-5.4"

Any OpenRouter-compatible model ID is accepted.

What the bot reads

For each mention, Critique assembles a context package that includes:

  • The full PR diff (up to 40 changed files)
  • Relevant repository files beyond the diff (fetched via codebase indexing)
  • The PR title and description
  • The full comment thread history
  • For inline comments: the file path, diff hunk, and line number

This context is assembled fresh for each mention — Critique does not maintain session state between separate PR comments.

Rate limits and scaling

There is no hard limit on the number of @critique mentions per PR. Each mention triggers an independent context assembly and model call. For very active PRs with long threads, full context is included each time — responses may be slightly slower as the thread grows.

The /review command triggers a complete pipeline re-run (Scout + Specialists + Lead), which consumes credits like an automated review. Plain mentions and other commands use a lightweight single-model call.