Claude Code Action
https://github.com/anthropics/claude-code-action- Category
- Developer Tools
- Rank
- No. 240Tools index
- Pricing
- Open Source
- Type
- TOOL
- Builder
- anthropics
- GitHub
- 8.8k stars
- Latest release
- v1.0.218
- Date
About
Official GitHub Action for Claude Code — wire Claude into your CI/CD to review PRs, run code reviews, or kick off agent workflows from GitHub events.
What it does
Claude works inside pull requests and issues as either an interactive collaborator or a prompt-driven automation worker. It reads repository context, answers questions, posts feedback, edits code, commits changes, and reports progress. A supplied prompt starts automation mode. Without one, mentions, assignments, or labels can start an interactive session.
Why it's ranked here
This is a capable repository agent with unusually broad workflow coverage and serious security controls. It supports interactive requests, scheduled maintenance, issue triage, structured outputs, CI log access, and several authentication providers. The main reservation is operational complexity: permissions, tool grants, event safety, credentials, and branch behavior all require deliberate configuration.
What's good
The action limits ordinary triggers to users with repository write access and blocks bots by default. Repository tokens stay scoped to the current repository. Bash commands require explicit permission, while web access is disabled by default in interactive mode. Workload identity federation avoids stored API keys, and validated JSON can feed later automation steps.
Tradeoffs
Mode detection is documented as experimental and may change or disappear. The official app cannot update workflow files, cross repository access is unavailable, and default operation neither creates pull requests nor performs rebases, merges, or force pushes. Public repository workflows remain exposed to prompt injection risks, especially when bot or non-writer restrictions are relaxed. Shallow clones may omit needed history.
How to use it well
Use it for teams already running GitHub Actions that want bounded code assistance, review checklists, issue triage, maintenance, or CI failure analysis. Grant only the GitHub and command permissions each workflow needs, keep credentials in secrets or use federation, and review generated changes before merging. Use local tooling when you need rebasing, destructive history changes, cross repository work, or automatic pull request creation.
Technical notes+
The private TypeScript package in package.json targets Bun for tests and scripts, with TypeScript type checking and dependencies on GitHub Actions, Octokit, the Claude Agent SDK, MCP, and Zod. src/modes/tag/index.ts prepares interactive runs, checks for a human actor, creates a tracking comment, configures branches and Git authentication, and grants a bounded tool set under acceptEdits. src/modes/agent/index.ts handles explicit-prompt automation without tracking comments and builds MCP configuration from allowed tools. src/create-prompt/index.ts validates supported GitHub event contexts and disables WebSearch and WebFetch by default. docs/security.md documents actor checks, repository-scoped tokens, prompt injection controls, protected base-branch configuration, and commit-signing options.
Observed
- License
- MIT License
- Primary language
- TypeScript
- Install surface
- Private package distributed for use as a GitHub Action
- Runtime and tooling
- Bun scripts, Bun tests, and TypeScript type checking
- Interface
- GitHub Action with workflow inputs, outputs, comments, and MCP tool integration
- Authentication providers
- Anthropic API or OAuth, workload identity federation, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry
- Execution platform
- Runs on the user's GitHub Actions runner
- Repository scope
- Each invocation is limited to the repository where it was triggered
Read from README.md, package.json, src/create-prompt/index.ts, src/modes/tag/index.ts, src/modes/agent/index.ts, docs/faq.md, docs/setup.md, docs/usage.md, docs/security.md, docs/solutions.md, docs/experimental.md, docs/configuration.md.
What it can do
Review pull requests automatically
GitHub pull request → Code review comments and feedback
Perform automated code reviews
Code changes in repository → Code quality analysis and suggestions
Trigger agent workflows from GitHub events
GitHub webhook events → Executed automated workflows
Integrate Claude AI into CI/CD pipeline
CI/CD configuration and GitHub repository → AI-powered automation in build process
Analyze code changes for issues
Git commits and diffs → Issue identification and recommendations
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.