Skip to content
Claudekit.ai

Claude Code Action: Claude in GitHub PRs and Issues

Claude Code Action runs Claude Code in GitHub Actions to answer @claude mentions, review PRs and implement changes. Setup, workflow example, pros and cons.

By Claudekit Team Published 1 min read

Type
Extension
Pricing
Open Source
Repository
github.com
GitHub stars
9,625
Website
github.com

Claude Code Action is Anthropic’s official GitHub Action. It runs Claude Code on your GitHub runner so Claude can answer questions, review pull requests, and implement changes from GitHub itself. It is technically a GitHub Action rather than an editor extension; we list it under extensions because it extends GitHub with Claude.

Setup

The quickest route, if you use the Anthropic API directly and are a repository admin: open claude in your terminal and run

/install-github-app

It walks you through installing the GitHub app and adding the required secrets. For Amazon Bedrock, Google Vertex AI or Microsoft Foundry, follow the cloud providers doc in the repo instead.

Workflow example

A minimal .github/workflows/claude.yml, adapted from the repo’s usage guide:

name: Claude Assistant
on:
  issue_comment:
    types: [created]
  pull_request_review_comment:
    types: [created]
  issues:
    types: [opened, assigned, labeled]

jobs:
  claude-response:
    runs-on: ubuntu-latest
    steps:
      - uses: anthropics/claude-code-action@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          # or: claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}

Then comment @claude on an issue or PR. Optional inputs include prompt for automation runs, claude_args to pass CLI flags, plugins to install Claude Code plugins, and trigger_phrase to change the mention.

Use cases

  • Automatic PR review, including path-specific or security-focused checklists.
  • Issue triage and labeling.
  • Small fixes implemented straight from an issue comment.
  • Scheduled maintenance and documentation sync jobs.

Pros

  • Official, open source, runs on your own runners.
  • Same configuration model as Claude Code: you can load plugins and MCP servers.
  • Works with several model providers.

Cons

  • Every run costs model usage; noisy triggers add up.
  • Write permissions on a repo require careful review of what Claude can push.
  • Setup beyond the quickstart (cloud providers, custom apps) takes some YAML work.

Want Claude to reach GitHub from your local sessions too? Add the GitHub MCP Server.

Frequently asked questions

Is Claude Code Action free?

The action itself is open source (MIT). Model usage is billed through whichever provider you authenticate with: an Anthropic API key, a Claude Code OAuth token, Amazon Bedrock, Google Vertex AI or Microsoft Foundry.

Where does the code run?

On your own GitHub Actions runner. Only the model API calls go to your chosen provider.

Can I change the @claude trigger?

Yes. The trigger_phrase input sets a custom phrase, and assignee_trigger or label_trigger let issues assigned or labeled a certain way start Claude.

Get the weekly Claude digest

New skills, MCP servers, prompts and Claude updates — one short email a week. No spam, unsubscribe anytime.

The newsletter is launching soon. Follow the RSS feed in the meantime.