cft0808/edict

Edict (also called "Sansheng Liubu", Chinese for "Three Ministries, Six Departments") is an open-source multi-agent AI orchestration system written in Python. It lets you coordinate a team of twelve specialized AI agents that work together to complete complex tasks, modeled after the imperial bureaucratic structure of Tang Dynasty China, where each "agent" plays a role like a crown prince, a planning ministry, a review board, and specialist departments for execution. The key idea is a mandatory quality-review layer built into the system's architecture: before any plan reaches the execution stage, a dedicated review agent must approve it. If the output doesn't meet the required standard, it's sent back, not just flagged as a warning. This distinguishes it from other multi-agent frameworks where agents produce results without any quality gate. The system includes a real-time web dashboard (a visual task board styled like a Kanban board, a board that organizes tasks by status: to-do, in-progress, done), full audit trails of every decision, the ability to pause or cancel tasks mid-run, and per-agent AI model switching without restarting the whole system. The backend runs on Python's standard library with no extra dependencies, the frontend is built with React 18. You can try it instantly via Docker or install it fully if you have Python 3.10 or later and the OpenClaw platform installed.