Collaborative multi-agent systems are a major frontier in AI, but they are often slow or expensive to run. The open-source project Auto-Company offers a practical alternative: a fully autonomous AI company that operates 24/7 directly on your local machine (Windows/WSL or macOS). Instead of a single model, it coordinates a virtual workforce of 14 domain-specific AI experts.
The 14-Agent Virtual Roster
- Leadership & Design: A CEO persona (modeled after Jeff Bezos) sets direction, supported by a CTO (Werner Vogels) and a Product Designer (Don Norman) to shape the product roadmap.
- Engineering & Quality: Specialized backend, frontend, and QA agents write code, debug issues, and manage automated tests.
- Growth & Admin: Dedicated marketing and operation agents compile marketing copy and track performance.
Key Architectural Components
- Claude Code & Codex Integration: The agents leverage Claude Code to interact directly with the local command line—editing source files, running compiler scripts, and deploying outputs.
- Daemon Execution Mode: Background service daemons for macOS and WSL2 keep the agent loop running continuously without blocking your terminal.
- Monitoring Dashboard: A lightweight local UI lets you observe agent logs, task boards, and communication in real-time.
Installation & Setup
- Step 1 (Clone Repo): Get the codebase:
git clone https://github.com/MaxMiksa/Auto-Company. - Step 2 (API Keys): Configure your API credentials for Anthropic (Claude) and Codex.
- Step 3 (Platform Install): Run the platform-specific installer script (ensure WSL2 is active on Windows).
- Step 4 (Launch Company): Initialize the continuous loop via:
npm run start:company.
Link to repo: MaxMiksa/Auto-Company
Agentic Systems for Your Enterprise?
We help you implement autonomous systems safely and efficiently — from architecture to deployment.
Request Consultation →