</>MCP Agents Market
Agent

Auto-Company

by MaxMiksa2.9kPythonUpdated 2026-05-20

An auto-company works for 24/7 on your own PC - Windows/Linux/macOS.

Claude CodeCodex

Auto-Company is a fully autonomous multi-agent AI system that orchestrates 14 specialized AI agents operating continuously on Windows, Linux, and macOS. Each agent embodies the expertise of a real-world industry leader—from Jeff Bezos-style CEO decision-making to DHH-inspired full-stack development—collaborating through structured workflows to ideate products, write code, deploy applications, and execute marketing without human intervention. The system runs as a daemon, cycling through autonomous work sessions where agents read shared consensus memory, form dynamic squads of 3-5 members, execute tasks, and hand off state to the next cycle. Powered by Claude Code or Codex CLI, it includes built-in guardrails, a local web dashboard for monitoring, and file-based steering that lets humans redirect the AI team by editing a single markdown file.

Key Features

14 expert-persona agents modeled after industry leaders (Bezos, DHH, Charlie Munger, Don Norman, etc.) with domain-specific mental models
24/7 autonomous operation via daemon mode with automatic crash recovery on macOS (launchd) and Windows/WSL (systemd)
Dynamic squad formation that selects 2-5 optimal agents per cycle based on current objectives
Cross-cycle consensus memory stored in a single markdown file that serves as the handoff baton between work sessions
Forced convergence workflows preventing endless discussion (ideation → validation → execution)
30+ reusable skills arsenal covering research, security audits, financial modeling, SEO, and UX analysis
Built-in safety guardrails preventing repo deletion, force pushes, credential leaks, and system directory modification
Local web dashboard for real-time monitoring of cycle status, costs, and agent activity with full chain-of-thought logs

Use Cases

  • 01Running a fully autonomous startup that ideates, validates, builds, and launches products without human coding
  • 02Continuous product development where AI teams handle feature development from design through QA and deployment
  • 03Autonomous market research and opportunity discovery combining competitive analysis with business validation
  • 04Hands-off DevOps and SRE operations with automated monitoring, deployment, and reliability management
  • 0524/7 business operations including pricing strategy, sales funnel optimization, and weekly performance reviews
  • 06Exploratory software prototyping where AI teams experiment with ideas and deliver working code continuously

Related Agents

View more

Auto-Company — FAQ

What is Auto-Company and how does it work?+

Auto-Company is a multi-agent AI system that runs 14 autonomous expert agents in continuous cycles. Each cycle, agents read shared memory, form a task-specific squad, execute work (research, coding, deployment, marketing), update consensus memory, then sleep briefly before the next cycle begins.

How do I install and start Auto-Company on macOS?+

Clone the repository, ensure Claude Code or Codex CLI is installed and authenticated, then run 'make start' for foreground execution or 'make install' to set up a daemon that auto-starts and auto-restarts. Prerequisites include macOS, Node.js, and available model quota.

How do I install Auto-Company on Windows?+

You need Windows 10/11 with WSL2 (Ubuntu) installed. Inside WSL, install Claude Code or Codex CLI, node, jq, and make. Clone the repo, then from PowerShell run '.\scripts\windows\start-win.ps1' to launch the daemon, which runs inside WSL but is controlled from Windows.

Which AI clients or engines does Auto-Company support?+

Auto-Company supports Claude Code (default) and Codex CLI as execution engines. You select the engine at startup with the ENGINE environment variable or -Engine parameter. Both must be installed and authenticated with available API quota.

Does Auto-Company require API keys and is it free to run?+

You need an authenticated installation of Claude Code or Codex CLI with available model quota. Each autonomous cycle consumes API credits, so ongoing operation incurs costs based on your provider's pricing. The software itself is MIT-licensed and free to use.

How do I control or redirect the autonomous AI team?+

Edit the 'Next Action' field in memories/consensus.md to steer the team's direction. Use 'make pause' and 'make resume' (or the Windows PowerShell equivalents) to halt or restart the daemon. The local dashboard accessible via 'make dashboard' provides real-time monitoring of agent activity and costs.

How do I install Auto-Company?+

Open the source repository on GitHub and follow its README. Auto-Company is a agent — MCP Agents Market links you directly to the official repo.

Is Auto-Company free?+

Auto-Company is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.

Related searches