This guide walks you through installing CRH and running your first AI code review against a Launchpad merge proposal. Before you begin, make sure you have Node.js 20 or later installed and a Launchpad merge proposal URL ready to test with.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/goulinkh/code-review-harness/llms.txt
Use this file to discover all available pages before exploring further.
Install CRH
Install the Verify the installation:
code-review-harness package globally. This provides both the crh and code-review-harness binary names.Run a dry-run review to stdout
Run a review against your merge proposal and print the results to stdout. This requires no credentials — it is a safe way to see CRH in action before posting anything to Launchpad.CRH uses The review findings are written as a single JSON line to stdout. Progress and agent thinking stream to stderr.
anthropic:claude-sonnet-4-6 by default. To use a different model, pass --model <provider>:<model-id>. For example:(Optional) Post the review to Launchpad
To post inline comments and a verdict back to the merge proposal, set your Launchpad OAuth credentials and switch the sink to Then run the review with
launchpad.--sink launchpad:The Launchpad sink requires valid OAuth 1.0a tokens. Without them, the request will be rejected by the Launchpad API. See the authentication guide for instructions on obtaining tokens.
Next steps
- Authentication guide — obtain and configure Launchpad OAuth tokens
- Model selection — switch LLM backends and use OpenAI-compatible servers
- Flags reference — full list of
crh reviewoptions