A 14-year-old developer has released Autark, an open-source tool that enables AI agents to coordinate tasks without requiring constant user prompting, according to the published report The system treats agents as a virtual team, allowing them to wake up via schedules, timers, or events and plan their own next steps.reported This shift moves beyond simple prompt-response interactions toward autonomous workflow management.Autark is a new tool that lets AI programs work together as a team, handling tasks automatically without needing you to type instructions for every step.Autark implements a manager-agent architecture that orchestrates task delegation and result reporting across model-agnostic AI agents, operating on event-driven triggers rather than synchronous prompts.
Evidence: reported
14-year-old creates tool for AI agents to band together like a teamaifleet-register
The figures
The Register reports that Autark was built by Stanislaus Hees, a 14-year-old Austrian developer.reported Autark integrates with OpenAI Codex, Claude Code, Gemini CLI, GitHub Copilot CLI, and Kimi Code. Local models are also supported. The Register reports that support for models other than Codex and Claude Code has not yet been tested in live runs.reported
Evidence: reported
14-year-old creates tool for AI agents to band together like a teamaifleet-register
Behind the figures
Hees developed Autark after finding existing tools insufficient for ongoing work organization. He stated that while he could get help with individual tasks using Claude Code and Codex, he still had to organize the work and track unfinished items. He sought a platform capable of managing continuous projects instead of merely replying to individual inputs. He used AI coding tools to help develop the framework itself.
What it means for your stack
Autark is model-agnostic and connects to various services through their CLI tools. A manager agent assigns tasks from an inbox to other agents for execution. These agents carry out the work and report results. If tasks are vague, agents can ask clarifying questions. Agents are activated by schedules, timers, new tasks, messages, or events rather than waiting for prompts.
Building on this
Developers can build multi-agent workflows where a manager agent delegates tasks to specialized agents. The documented interface allows agents to report results and ask for clarification when tasks are ambiguous. Because Autark is model-agnostic, engineers can integrate local models or connect to commercial CLIs like Codex or Claude Code. The tool requires users to establish their own security boundaries.
Why it matters
- Inference engineers can automate complex, multi-step workflows by using Autark’s manager-agent architecture to delegate tasks without manual intervention.
- The model-agnostic design allows teams to test different LLMs, though only Codex and Claude Code have confirmed live-run testing, creating potential integration risks for other models.
- Security responsibility shifts entirely to the user; developers must implement strict boundaries to prevent agents from running roughshod over systems with unfettered access.
- The tool’s ability to wake via events or schedules enables building persistent, autonomous agents rather than ephemeral prompt-response bots.
Topics: AI agents, open-source, automation, multi-agent systems, developer tools
Updated
Then / now / next
Developers had to manually organize work and track unfinished tasks even when using AI coding assistants like Claude Code or Codex.
Autark allows AI agents to act autonomously as a virtual team, delegating tasks and planning next steps without constant prompting.
Observers should watch for expanded testing of Autark’s support for models beyond Codex and Claude Code, as current support is untested in live runs.
Story timeline
- Then Developers had to manually organize work and track unfinished tasks even when using AI coding assistants like Claude Code or Codex.
- 10 Nov 2026 14-year-old creates tool for AI agents to band together like a team
- Now Autark allows AI agents to act autonomously as a virtual team, delegating tasks and planning next steps without constant prompting.
- Next Observers should watch for expanded testing of Autark’s support for models beyond Codex and Claude Code, as current support is untested in live runs.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
Model-agnostic
A system that can work with various large language models rather than being tied to one specific provider.
CLI
Command-line interface, a text-based interface used to interact with software tools and services.
Manager agent
An AI agent that oversees and delegates tasks to other agents within a multi-agent system.
Autark
An open-source tool that enables AI agents to coordinate tasks autonomously without constant user prompting.
Questions readers ask
Which AI models does Autark support?
Autark is compatible with OpenAI Codex, Claude Code, Gemini CLI, GitHub Copilot CLI, Kimi Code, and local models, but live testing has only been conducted with Codex and Claude Code.
How do agents in Autark initiate tasks?
Agents are activated by schedules, timers, new tasks, messages, or events, enabling them to operate without requiring constant user prompts.
Who developed Autark and why?
Teenage Austrian developer Stanislaus Hees built Autark to manage continuous tasks rather than merely reacting to single requests, a need he identified after finding existing AI tools insufficient for organizing ongoing work.
Connected stories
Reading receipt
- Autark allows AI agents to act on schedules or events without constant user prompting, functioning as a virtual team.
- Autark is model agnostic and can be configured to link with OpenAI Codex, Claude Code, Gemini CLI, GitHub Copilot CLI, and Kimi Code through their respective CLI tools.
- The tool’s website notes that while support for models beyond Codex and Claude Code is included, it has not yet been tested in live runs.
Still open: How does Autark handle security boundaries when given unfettered access to a PC?
Still open: What specific limitations exist for local model support in the current version?