Quick Start
Up and running with CHAOS-AI in under 5 minutes.
Run the PM Agent
The PM (Project Manager) agent is your entry point. Navigate to your project directory and run:
cd your-project
cw run pm
The PM agent will load your project context, display available agents, and wait for instructions.
Your First Task
Ask the PM to analyze your codebase:
> analyze the project structure and suggest improvements
The PM dispatches the appropriate agents in parallel — typically review, test, and docs agents — and coordinates their findings into a single structured report.
Run a Specific Agent Directly
Skip the PM and run any agent directly:
# Python code generation
cw run python-agent "add input validation to the user model"
# Generate tests for a directory
cw run test-agent "generate tests for src/auth/"
# Security audit
cw run security-agent
# Generate documentation
cw run docs-agent "update README and API docs"
Connect Your Editor via MCP
Expose all 134 CHAOS tools to Claude Desktop, VS Code Copilot, Cursor, or Windsurf:
# stdio transport (for Claude Desktop / local clients)
cw mcp
# HTTP transport (for persistent hosting)
cw mcp --transport streamable-http
Then add CHAOS as an MCP server in your editor's config. See MCP Setup for client-specific instructions.
Run the Web Dashboard
Start the full stack with Docker:
docker compose up -d
The dashboard at http://localhost:3000 gives you live agent status, session history, token analytics, and the event stream.
Common Workflows
# Pre-commit: test + security + review (parallel)
cw run pm "run pre-commit pipeline"
# Feature complete: refactor + test + docs + review
cw run pm "feature complete pipeline for src/payments/"
# Project audit
cw run project-audit
# Security scan
cw run security-agent
Next Steps
- Configuration — Customize CHAOS for your project
- Agents — Browse the full agent catalog
- MCP — Connect your editor
- Use Cases — See what teams build with CHAOS