Bugzy

Open-source AI agent configuration for QA automation

Bugzy is a CLI tool that manages AI agent configuration for your AI coding assistant, providing properly configured subagents, tasks, and MCP servers for test automation.

Supported Tools

Note: Cursor and Codex support is experimental. Some features may not work as expected. Claude Code is the recommended and fully tested option.

Features

• ✅ Complete Task Library - 8 pre-built QA automation tasks
• ✅ Flexible Subagents - Test Runner, Test Code Generator, Test Debugger & Fixer, Team Communicator, Documentation Researcher, Issue Tracker
• ✅ Easy Setup - Interactive CLI configuration (bugzy setup)
• ✅ Local Execution - Runs entirely on your machine with Claude Code
• ✅ Version Control Friendly - Configuration stored in git-friendly files
• ✅ Language Agnostic - Works with any project type (Python, Ruby, Go, JavaScript, etc.)
• ✅ No Vendor Lock-in - Full control over customization
• ✅ MCP Integration - Automatic MCP server configuration

Quick Start

# Install globally
npm install -g @bugzy-ai/bugzy

# Navigate to your project
cd my-project

# Run interactive setup
bugzy setup

# Configure your MCP secrets
cp .env.example .env
vim .env  # Add your MCP API tokens (Slack, Notion, etc.)

# Generate test plan (creates .env.testdata with test data)
bugzy "/generate-test-plan for [your feature]"

# Start a Claude Code session
bugzy

Example Usage

# Generate a test plan
bugzy "generate test plan for user authentication"

# Or use slash commands in Claude Code
/generate-test-plan for checkout flow
/run-tests for payment processing
/verify-changes deployed new feature

What Gets Created

After running bugzy setup, your project will have:

your-project/
├── .bugzy/
│   ├── config.json              # Your subagent configuration (includes selected tool)
│   └── runtime/
│       ├── project-context.md   # Project information
│       └── templates/           # Customizable templates
├── .<tool>/                     # Tool-specific directory (.claude/, .cursor/, or .codex/)
│   ├── commands/                # Task commands (or prompts/ for Codex)
│   ├── agents/                  # Configured subagent prompts
│   └── mcp.json                 # MCP server configuration
├── .env.example                 # MCP secrets template (empty values)
├── .env.testdata                # Test data with actual values (from /generate-test-plan)
└── .env                         # Your actual secrets (gitignored)

Note: The directory structure varies by tool. Claude Code uses .claude/, Cursor uses .cursor/, and Codex uses .codex/.

Available Subagents

Available Tasks

Configuration

Bugzy uses a simple configuration format:

{
  "version": "1.0.0",
  "tool": "claude-code",
  "project": {
    "name": "my-project"
  },
  "subagents": {
    "test-runner": "playwright",
    "team-communicator": "slack",
    "documentation-researcher": "notion",
    "issue-tracker": "jira-server"
  }
}

The tool field can be "claude-code" (default), "cursor" (experimental), or "codex" (experimental).

Don't edit config.json manually - use bugzy setup to reconfigure.

Note: When using CLI (bugzy), the team-communicator is automatically configured to use terminal-based communication. No setup required.

Environment Variables

Bugzy loads environment variables for MCP servers and test configuration:

# MCP Secrets
SLACK_BOT_TOKEN=xoxb-...
NOTION_TOKEN=secret_...
RESEND_API_KEY=re_...

# Test Configuration
TEST_BASE_URL=http://localhost:3000
TEST_USER_EMAIL=test@example.com
TEST_USER_PASSWORD=secure-password

See Configuration Guide for complete details.

Documentation

• 📚 Getting Started Guide - Installation and first-time setup
• ⚙️ Configuration Guide - Subagent and environment setup
• 🤖 Subagents Guide - Available subagents and integrations
• 📋 Task Library - All available tasks and usage
• 🏗️ Architecture - How Bugzy works under the hood

Examples

Check out example configurations:

• Basic Setup - Minimal configuration with just Test Runner
• Full Setup - All subagents configured

Requirements

• Node.js v18 or higher
• AI Coding Tool: One of the following installed and configured:
  • Claude Code (recommended) - Anthropic's official CLI
  • Cursor (experimental) - VS Code-based AI editor
  • Codex CLI (experimental) - OpenAI's terminal-based agent
• npm or yarn package manager

Breaking Changes (v0.2.0)

New Test Execution Output Format

This is a breaking change. Bugzy now uses a hierarchical test execution format with custom Playwright reporter:

What's New:

• Custom Bugzy Reporter: Automatically creates test-runs/YYYYMMDD-HHMMSS/ structure with manifest.json
• Execution Retries: Tracks multiple attempts per test (exec-1/, exec-2/, exec-3/)
• Comprehensive Artifacts: Videos for all tests, traces/screenshots for failures only
• Manifest Format: New manifest.json provides complete test run summary and per-test execution details

Migration Path:

1. Update to v0.2.0: npm update -g bugzy
2. Run bugzy setup in your project to regenerate configuration files
3. New files will be created:
   • playwright.config.ts - Uses custom Bugzy reporter
   • reporters/bugzy-reporter.ts - Custom reporter implementation
   • .bugzy/runtime/templates/test-result-schema.md - Complete schema documentation

Key Changes:

• /run-tests now reads manifest.json instead of .last-run.json
• Test artifacts organized in test-runs/{timestamp}/{testId}/exec-{num}/ structure
• Environment variable BUGZY_EXECUTION_NUM controls retry attempts
• Videos recorded for ALL tests (not just failures)
• Trace/screenshots only for failures (more efficient)

See .bugzy/runtime/templates/test-result-schema.md for complete format documentation.

Reconfiguration

Need to change your setup? Just run:

bugzy setup

Bugzy will detect your existing configuration and allow you to make changes.

Troubleshooting

Command Not Found

npm install -g @bugzy-ai/bugzy

Missing Secrets

✗ Missing required MCP secrets: SLACK_BOT_TOKEN

Add the missing secret to your .env file.

AI Tool Not Found

Install your selected AI coding tool and ensure it's in your PATH:

• Claude Code: Install from claude.ai/code
• Cursor: Install from cursor.com
• Codex CLI: Install from OpenAI

See the Getting Started Guide for more troubleshooting tips.

FAQ

Q: Do I need a package.json? A: No! Bugzy works with any project type.

Q: Can I use this with Python/Ruby/Go projects? A: Yes! Bugzy is language-agnostic.

Q: Where are my secrets stored? A: In .env which is gitignored. .env.example (MCP secrets template) and .env.testdata (test data) are committed.

Q: Can I customize the tasks? A: You can customize templates in .bugzy/runtime/templates/.

Q: How do I update to get new tasks? A: Run npm update -g @bugzy-ai/bugzy && bugzy setup

License

MIT © Bugzy Team

Contributing

Contributions are welcome! Please read our contributing guidelines (coming soon).

Support

• Issues: github.com/bugzy-ai/bugzy/issues
• Discussions: github.com/bugzy-ai/bugzy/discussions
• Documentation: docs.bugzy.dev