Claude Code Action runs Claude Code inside GitHub Actions workflows so it can answer questions and make code changes on pull requests and issues. It decides when to activate from the workflow context: an @claude mention in a comment, an issue assignment, or an automation task given an explicit prompt.
The action executes on your own GitHub runner, and model calls go to the provider you choose: the Anthropic API (API key or workload identity federation), Amazon Bedrock, Google Vertex AI, or Microsoft Foundry. Configuration uses a single prompt input plus pass-through Claude Code arguments.
Documented automation patterns include automatic pull request review, reviews triggered by changes to specific paths, extra handling for external contributors, custom review checklists, scheduled repository maintenance, issue triage and labeling, documentation sync, and security-focused reviews aligned with OWASP.
Features
- Mode detection: picks the execution mode from the workflow context (mention, assignment, or explicit prompt) without extra configuration
- Code questions: answers questions about code, architecture, and programming in pull request and issue threads
- Code review: analyses pull request changes and suggests improvements
- Code implementation: makes simple fixes, refactors, and new features directly on the branch
- Tool access: GitHub APIs and file operations by default, with further tools and MCP servers enabled through configuration
- Progress tracking: a comment with checkboxes that update as tasks complete
- Structured outputs: validated JSON results exposed as GitHub Action outputs for downstream steps
- Provider choice: Anthropic API, Amazon Bedrock, Google Vertex AI, or Microsoft Foundry
