14-year-old creates tool for AI agents to band together like a team
Autark can act without constant prompting, but leaves users to enforce their own security boundaries
A 14-year-old Austrian developer named Stanislaus Hees has created a new tool called Autark that allows AI agents to work together like a team without constant prompting. Autark treats AI agents as a virtual team, with a manager agent delegating tasks to other agents, who then complete them and report their results. If tasks are vague, agents can ask clarifying questions.
Autark is model-agnostic and can connect to various AI services like OpenAI Codex, Claude Code, Gemini CLI, GitHub Copilot CLI, and Kimi Code, as well as local models. However, Autark's capabilities come with potential risks. It doesn't have any sandboxing, approval prompts, or permission gates, so agents run with the privileges of the machine user.
This means they could potentially access sensitive information, delete files, or send unsolicited emails if given the wrong instructions. In live test runs, agents even accessed real repositories, backups, SSH keys, and .env files within minutes. Hees developed Autark to address his own frustration with existing AI tools, which required him to organize work and keep track of unfinished tasks.
The tool is still under active development and is available for self-hosting under an Apache open-source license.
Written by urgent.news from The Register Software's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
