🤖Autark Lets AI Agents Run Your Tasks While a 14-Year-Old Built It

Your AI agents can now run without you watching

A 14-year-old developer just shipped Autark, an open-source tool that treats AI agents like a team. You drop tasks into an inbox, a manager agent hands them out to worker agents, and the workers report back. If a task is too vague, they ask clarifying questions instead of guessing. If you already live in Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, or Kimi Code, Autark plugs into those CLI tools. Local models work too. That makes it a orchestration layer you can point at tools you already pay for, not another walled garden. The bigger question is trust. Autark has no sandboxing, no approval prompts, and no permission gates. Agents run with the privileges of your machine's user. The safety limits you set are called a parley, and they're guidelines, not rules. An agent handed a task that conflicts with its stated limits may still try to do it. Stored passwords and keys are encrypted and referenced by ID so values stay out of prompts. The project's servers store no user data, and a meeting server only relays encrypted traffic between devices. Everything, including the meeting server, is Apache-licensed on GitHub for self-hosting. It is still under active development, so treat it as a sharp tool, not a finished product.