Autoplay

Autoplay lets you click a button and have AI automatically implement an issue, run tests, and create a pull request. It uses Claude Code (Anthropic) in GitHub Actions.

How It Works

1. Click "Autoplay" on an issue in SprintFlint
2. SprintFlint triggers a GitHub Actions workflow in your repository
3. Claude Code implements the issue
4. Tests are run automatically to verify the implementation
5. A pull request is created when everything passes
6. The PR links back to the original issue

Repository Setup

your-repo/
├── .github/workflows/autoplay.yml   # GitHub Actions workflow
├── Procfile.autoplay                # Foreman config for CI (optional)
└── ...

Claude Code Setup

Use the official claude-code-action GitHub Action:

- name: Run Claude Code
  uses: anthropics/claude-code-action@v1
  with:
    claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
    # Or: anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
    prompt: |
      Implement issue ${{ inputs.issue_number }}: ${{ inputs.issue_title }}
      ${{ inputs.issue_description }}
      Create branch '${{ inputs.branch_name }}', commit, and open a PR.

      If you have questions or cannot complete the issue, use the SprintFlint API
      to add a comment: POST /api/v1/projects/.../issues/.../comments with
      X-API-Token header.
    github_token: ${{ github.token }}

Claude Code Authentication

• OAuth (Pro/Max): Run claude setup-token, then add CLAUDE_CODE_OAUTH_TOKEN
• API key: From console.anthropic.com, add ANTHROPIC_API_KEY
• Bedrock/Vertex: use use_bedrock: true or use_vertex: true in the action

SprintFlint API Integration

The AI can use the SprintFlint REST API to leave comments when it has questions or needs clarification:

curl -X POST https://sprintflint.com/api/v1/projects/{project_id}/sprints/{sprint_id}/issues/{issue_id}/comments \
  -H "X-API-Token: $SPRINTFLINT_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"comment": {"body": "Question: Should this handle edge case X?"}}'

SprintFlint API Token

Required for AI feedback. Lets the AI leave comments when it has questions.

1. Go to your Profile
2. Open the "API Token" section and generate a token
3. Add as repository secret: SPRINTFLINT_API_TOKEN

Browser Testing

Install agent-browser for UI testing. Claude Code can use it via shell commands:

- name: Install agent-browser
  run: |
    npm install -g agent-browser
    agent-browser install

agent-browser open http://localhost:3000
agent-browser snapshot -i    # refs @e1, @e2, ...
agent-browser click @e1
agent-browser fill @e2 "text"
agent-browser screenshot page.png

Project-Specific Setup

Add the steps needed to run your stack before the AI step. Example for Rails:

- uses: ruby/setup-ruby@v1
  with:
    ruby-version: '3.4.7'
    bundler-cache: true
- uses: actions/setup-node@v4
  with:
    node-version: '24.x'
    cache: 'yarn'
- name: Setup database
  run: bundle exec rake db:create db:migrate db:seed
  env:
    RAILS_ENV: test

Troubleshooting

Workflow not triggering

• Ensure autoplay.yml is in .github/workflows/ on the default branch
• Check workflow syntax (YAML and step names)

Authentication errors

• Check CLAUDE_CODE_OAUTH_TOKEN or ANTHROPIC_API_KEY
• Regenerate tokens if they may have expired

Tests failing

• Inspect the workflow logs in GitHub Actions
• Ensure project-specific setup (DB, env, assets) is complete and matches local

Next Steps

• API Reference - Integrate with your own tools
• GitHub Integration - Learn about PR linking