Skip to main content

agentorg

AI Agents
Open Source
A framework for generating multi-agent organisations
Published

November 4, 2026

agentorg is a framework for generating multi-agent organisations from one Python file, spec.py. It’s still early and isn’t on PyPI yet, so install it from GitHub:

pip install git+https://github.com/georgedouzas/agentorg.git
agentorg --help
Usage: agentorg [OPTIONS] COMMAND [ARGS]...

  Agentorg: Generate runnable multi-agent organisations.

Options:
  --version  Show the version and exit.
  --help     Show this message and exit.

Commands:
  new     Interview the user to assemble an agentorg spec.py.
  render  Render a project from a specification file.
  ui      Run the browser-based master interview UI.

agentorg new starts an interview with a master agent, which fills in the spec for you. There is also agentorg ui if you prefer a browser. The interview moves through five stages: discover, configure, build, refine and handoff. Each stage’s agent only has the tools for that stage, so the discovery agent can’t set your organisation’s identity by accident.

agentorg render then turns the spec into a project. By default that’s the full thing, with a backend and a React dashboard. You can also render just the frontend, backend or AI layer.

The master agent uses Agno, but your own agents, teams and workflows don’t have to. In the spec, each one can be anything runnable, like a plain async function.