Resources
FAQ
Common questions about Codebot.
Which git hosts does Codebot support?
GitHub and Azure DevOps. Reviews, checks, re-reviews and mentions work on both. GitLab and Bitbucket aren’t supported yet.
Which models can I use?
Any model from a provider you add a key for: Anthropic, OpenAI, Google Gemini, OpenRouter, Groq, Mistral, DeepSeek, or any OpenAI-compatible endpoint. You choose one model for reviews and one for the agent. See Model providers.
Do I pay for tokens through Codebot?
No. Runs use your organization’s keys, so your model provider bills you directly. Codebot counts compute: the time runs spend in cloud sandboxes. See Usage and limits.
Does Codebot run our tests?
Not in reviews, which are read-only so they stay fast and safe. The agent can run commands in its sandbox, such as your tests, when it works on a request.
Can Codebot block a merge?
Yes, if you turn on Fail the check on blocking issues and require its check. It only blocks on open critical or high findings, and never because a run failed. See Checks and merge gate.
How do I stop Codebot commenting on something?
Reply on the thread explaining why, or resolve it; the next review takes it into account. For a whole class of comments, add a rule to the repository’s .codebot/review.md, the review Instructions, or raise the Inline comments threshold.
Can I use different settings per repository?
Not yet. Review settings, triggers and skip rules apply to every repository. Per-repository review rules go in a .codebot/review.md file.
Does Codebot store my code?
It clones your repository into a temporary sandbox for each run. Agent conversations keep a stopped sandbox for a while so follow-ups can resume, then it’s deleted. See Security and data.
How do I get access?
Codebot is in private beta. Request access and we’ll send you an invite.