Crew44 orchestrates specialist AI agents locally, enabling them to collaborate within a unified workspace. It targets developers and power users seeking to leverage multiple AI models concurrently for complex tasks, offering compounded skills and role-specific model optimization. The primary benefit is enhanced productivity and context retention compared to single-agent generalist approaches, all while maintaining user privacy and data control.

How It Works

Crew44 employs a local-first architecture comprising an Electron/React UI and a Go daemon. The daemon discovers and manages installed coding agent CLIs (e.g., Claude, Codex, Gemini), routing tasks to agents bound to specific models and skills. Skills are file-based capabilities (SKILL.md + assets) that agents can access, allowing for compounding knowledge across tasks and providers. Agents communicate via WebSocket JSON-RPC, handing off work through explicit "handovers" with brief descriptions, facilitating parallel specialist execution. State is stored as plain files, ensuring transparency and local control.

Quick Start & Requirements

• Installation: Download signed desktop builds (.dmg, .exe, .AppImage/.deb) from crew44.io/download, or build from source.
• Prerequisites: Node.js (20+), Go (1.26+), and at least one supported coding agent CLI installed (e.g., Claude, Codex, Cursor, Gemini).
• Setup: Running npm install and npm run dev initiates the build and launches the application.
• Links: Download: crew44.io/download, Website: crew44.io, Docs: README.

Highlighted Details

• Local-First & Privacy: All UI, state, and orchestration run on 127.0.0.1. No cloud accounts, subscriptions, or telemetry; network traffic is limited to underlying agent calls.
• Compounding Skills: Skills captured as SKILL.md are persistent and accessible by any agent, building a shared knowledge base.
• Model Specialization: Assigns the optimal model (e.g., Opus for planning, GPT-5.5 for coding) to specific agent roles, improving efficiency and cost-effectiveness.
• Mobile Pairing: An optional Expo mobile app allows remote monitoring and interaction via an end-to-end encrypted tunnel.

Maintenance & Community

The project is actively maintained, indicated by recent CI activity and a clear MIT license. Community links such as Discord or Slack are not explicitly detailed in the README.

Licensing & Compatibility

Licensed under the MIT License, permitting broad use, modification, and distribution, including for commercial purposes, with minimal restrictions.

Limitations & Caveats

The system relies on the user having pre-installed and configured coding agent CLIs. While the runtime layer is extensible via Go adapters, adding new agent types requires development effort. The mobile pairing feature uses a relay for NAT traversal, though it is end-to-end encrypted.