Skip to content

Getting Started

Assemble is an open-source prompt orchestration system that generates native configuration files for 21 AI coding platforms. It gives you a team of 34 specialized AI agents — coordinated by Jarvis, the orchestrator — without any runtime, daemon, or background process.

  1. You run npx cohesiumai-assemble in your project
  2. Assemble generates platform-native config files (.cursorrules, CLAUDE.md, .github/copilot-instructions.md, etc.)
  3. Your AI coding tool reads these files automatically
  4. You use commands like /go, /feature, /bugfix to orchestrate agents

Zero runtime. Zero dependency. Pure prompt engineering.

  • Node.js 18 or later
  • An AI coding tool: Cursor, Claude Code, VS Code with Copilot, Windsurf, etc.
Terminal window
npx cohesiumai-assemble

The interactive wizard guides you through:

  1. Selecting your target platforms (Cursor, Claude Code, VS Code, etc.)
  2. Choosing a team profile (startup, enterprise, or agency)
  3. Configuring governance level (none, standard, strict)
  4. Setting language preferences

After setup, Assemble creates platform-native configuration files in your project root:

your-project/
├── .cursorrules # Cursor
├── CLAUDE.md # Claude Code
├── .github/
│ └── copilot-instructions.md # GitHub Copilot
├── .windsurfrules # Windsurf
├── .assemble.yaml # Assemble configuration
├── .assemble/
│ ├── agents/ # Agent definitions
│ ├── workflows/ # Workflow definitions
│ └── rules/ # Routing, teams, orchestrator
└── assemble-output/ # Workflow deliverables

Once configured, open your AI coding tool and try:

/go fix the auth bug in the login flow

Jarvis (the orchestrator) will:

  1. Assess complexity (TRIVIAL, MODERATE, or COMPLEX)
  2. Select the right agents (e.g., @bruce-banner for backend, @hawkeye for tests)
  3. Chain them in the optimal order
  4. Produce deliverables in assemble-output/