Real bugs (logic errors, broken edge cases), security issues in the change (injection risks, auth gaps, committed secrets), race conditions, error-handling gaps, API misuse, and cross-file regressions — with committable suggested fixes. Reviews run when a PR is opened and on every push to it; plain commits outside a PR are not scanned.
Your code is never stored by CodeMouse and never used for training. Reviews run on your own API keys; OpenAI and Anthropic do not use API data for training by default, and OpenRouter privacy settings let you exclude providers that train on inputs.
Trigger a review manually by commenting on any pull request — no push needed. Useful after fixing a provider issue (e.g. topping up API credits):
@codemouse review — run a review now (incremental when a previous review exists)@codemouse full review — re-review the entire PR diff from scratch$10/month with a 14-day free trial. Plus your own Anthropic and/or OpenAI API usage.
CodeMouse is a Squidcode product.