so-agent

A controlled multi-agent execution runtime featuring elastic single/multi-agent fusion. Enforces "LLM thinks, code decides" via zero-trust permissions and shadow state machines, dynamically downgrading simple tasks to single-agent fast paths. Built on the OpenAI Agents SDK.

GitHub

Official source

Start locally with the CLI guide

Back to catalog
My interests내 관심 주제

Tools

Sort

Loading catalog…카탈로그를 불러오는 중…