A Bun-based CLI tool that uses AI-powered test generation and human-in-the-loop workflow to create and execute ephemeral testing scenarios from GitHub PRs and Jira tickets. TAP combines Claude CLI for intelligent test generation, Claude Code for human refinement, and CUA (Computer Use Agent) for automated test execution with Docker-based containerized testing.
📖 View the complete documentation and guides in our wiki
brew install --formula https://raw.githubusercontent.com/mzieniuk/tap/main/Formula/tap.rbcurl -fsSL https://raw.githubusercontent.com/mzieniuk/tap/main/install.sh | bash- Download the appropriate binary for your platform from releases
- Make it executable:
chmod +x tap-* - Move to PATH:
sudo mv tap-* /usr/local/bin/tap
tap setup # Interactive configuration for API credentials and app setup# Step 1: Generate AI scenarios and export context for review
tap generate-tests <pr-url>
# Step 2: Review and refine scenarios with Claude Code
cd ./test-pr-{PR-number}-{commit-sha}
./claude-refine.sh # Or manually review exported files
# Step 3: Execute refined scenarios
tap execute-scenarios --file ./refined-scenarios.json- Claude CLI:
npm install -g @anthropic-ai/claude-cli && claude auth - Python 3.10+: Required for CUA agent
- Docker: Required for containerized test execution
- macOS: Docker Desktop
- Linux:
sudo apt install docker.io - Windows: Docker Desktop with WSL2
- CUA (Computer Use Agent): Auto-installed during setup
TAP requires API credentials and app setup instructions.
Run tap setup for interactive configuration.
Or use environment variables:
GITHUB_TOKEN- GitHub Personal Access TokenATLASSIAN_API_TOKEN- Unified token for Jira and ConfluenceATLASSIAN_EMAIL- Atlassian account emailATLASSIAN_BASE_URL- Atlassian instance URLTAP_APP_SETUP_INSTRUCTIONS- Natural language app setup instructionsANTHROPIC_API_KEY- Required for CUA (Computer Use Agent)
- 📖 Complete guides and workflows
- 💡 Usage examples and configuration
- 🏗️ Development and build instructions
TAP creates ./test-pr-{PR-number}-{commit-sha}/ directories with:
- Context Export: PR analysis, Jira tickets, Confluence docs
- AI Scenarios: Machine and human-readable test scenarios
- Refinement Tools: Claude Code instructions and helper scripts
- Execution Results: Screenshots, videos, QA reports
Use --verbose for detailed logging and --output <path> for custom directories.