Before you start
- Know which schema version your file uses. Your editor, the CLI and the web appâs YAML editor check against the V2 schema.
- Know which branch your repository reads configuration from. See Choose the configuration branch.
- To check the result of any step, use Check which configuration a review used.
Validate in your editor
Put the schema comment on the first line of the file so editors that use the YAML language server autocomplete keys and flag mistakes..coderabbit.yaml
Validate with the CLI
Run the validator in a repository that has a.coderabbit.yaml or .coderabbit.yml file:
1 when validation fails. Add --json or --agent for machine-readable output, for example in CI.
The command downloads the schema from
www.coderabbit.ai, so allow that host on restricted networks. It needs CLI 0.7.1 or later and doesnât appear in --help. It validates against the V2 schema only, so a file with version or remote_config fails.Validate in the web interface
The YAML editor view of the settings page validates YAML as you type and blocks saving until you fix the errors. It checks against the V2 schema that ships with the web app, and an unknown key is an error, so aversion key is flagged there too. Organizations in the V3 limited preview get a separate V3 editor that checks against the V3 schema. A custom Configuration source is validated when you save it: if CodeRabbit canât read the branch or its YAML is invalid, the previous source stays active.
Validate in the browser
Paste a configuration into the editor below to check its YAML before you commit it.Type-check a TypeScript configuration
With the SDK installed and the file in your TypeScript project, run the compiler in CI:Fix version errors
Version errors and what they mean
Version errors and what they mean
CodeRabbit doesnât retry a file that fails V3 validation as V2. The file yields no settings, and the review uses the next layers. Fix the error or restore the V2 file.
Test in the pull request that changes the file
1
Change the file on a branch
Edit
.coderabbit.yaml and open a pull request.2
Read the review
Check Source mode and Commit in Run configuration first. The review uses your edit only for a trusted pull request under Auto, or under Default branch with Allow configuration changes from trusted PRs on. Otherwise, run a preview review. It needs a draft pull request from a branch in the same repository and write access, so ask a collaborator with write access when you donât have it. Then check the pull request summary for a parsing warning, and judge the review against what you changed.
3
Merge when it looks right
Other pull requests keep their current configuration until the change merges.
If the selected source canât be read, the review falls back to Auto and shows a warning.
Run a preview review from a draft pull request
Comment exactly this on a draft pull request:You need write access. The pull request must be a trusted draft from a branch in the same repository, and its configuration must parse. Help lists the command for GitHub and GitHub Enterprise Server only.
Compare before and after in the web interface
When you select Apply changes in a repositoryâs settings, the Review changes dialog opens before anything is saved.- Your changes lists each edited setting with its saved value, your edit and its impact.
- On GitHub Cloud, a Current and After submit toggle shows the effective YAML before and after the change, with a full YAML comparison and the configuration sources before and after. Other platforms show the list of changes only.
Try a setting on one review
Add a fenced YAML block after the command to apply it to a single review:inheritance: true, the block replaces the resolved repository configuration, and omitted keys take schema defaults. It canât change approval policy or risk instructions. See Change settings for one pull request.
Preview through the API
preview_repository_configuration resolves the effective YAML for settings you supply and saves nothing. See Automate configuration with the API, MCP and CLI.
Check the result
Comment@coderabbitai configuration on the pull request and confirm the layers and values you expect. See Check which configuration a review used.
Whatâs next
Check the configuration used
See which layer supplied each value.
Change one pull request
Override settings for a single review.
Fix settings not applied
Diagnose a change that has no effect.