CE

cft0808/edict

Developer tools

This page was translated automatically. Technical details may differ from the original. View the original

17K stars Quality 71 Trend 44MIT

🏛️ Three Departments and Six Ministries System · OpenClaw Multi-Agent Orchestration System — 9 specialized AI agents with real-time dashboard, model config, and full audit trails

Overview

I redesigned the AI multi-agent collaboration architecture using the imperial system from 1,300 years ago. The results showed that the ancients understood differentiation and checks and balances better than modern AI frameworks. Twelve AI Agents (11 business roles + 1 compatible role) form three ministries and six ministries: Crown Prince sorting, Central Secretariat planning, Chancellery reviewing and sealing, Ministry of State Affairs dispatching, and six ministries + Ministry of Personnel executing in parallel. One more layer of institutional review than CrewAI, and one more real-time dashboard than AutoGen. 🎬 Watch the demo · 🚀 30-second experience · 🏛️ Architecture · 📋 Kanban functionality · 📚 Architecture documentation · English · Japanese · Contributions and participation Your browser does not support video playback. Please check the GIF or download video below. 🎥 Demonstration of the full AI multi-agent collaboration process across three provinces and six departments 📸 GIF preview (faster loading) Feishu decree → Crown prince's selection → Planning by the Secretariat → Review by the Chancellery → Six Ministries executing simultaneously → Memorial report (30 seconds) 🐳 Run a single line of docker run -p 7891:7891 cft0808/edict to experience the full Kanban demo (preset simulation data). The Agent collaboration model between CrewAI and AutoGen is—no one checks output quality. It's like a company without a QA department, where engineers write code and go online. - 📋 —— Is the Central Secretariat's plan complete? Is the breakdown of subtasks reasonable? - 🚫 —— Not a warning, it's a direct redo - 🔄 —— Only released after the plan meets standards

Overview (machine translated)

⚔️ Three Departments and Six Ministries · Edict

I redesigned the AI multi-agent collaboration architecture using the imperial system from 1,300 years ago. The results showed that the ancients understood differentiation and checks and balances better than modern AI frameworks.

Twelve AI Agents (11 business roles + 1 compatible role) form three ministries and six ministries: Crown Prince sorting, Central Secretariat planning, Chancellery reviewing and sealing, Ministry of State Affairs dispatching, and six ministries + Ministry of Personnel executing in parallel. One more layer of institutional review than CrewAI, and one more real-time dashboard than AutoGen.

🎬 Watch the demo · 🚀 30-second experience · 🏛️ Structure · 📋 Board Features · 📚 Architecture documentation · English · Japanese · Participate in contributions


🎬 Demo

Your browser does not support video playback. Please check the GIF below or download the video.

🎥 Full-process demonstration of AI multi-agent collaboration across three provinces and six ministries

📸 GIF Preview (Faster Loading)

Feishu issued imperial edicts → Crown Prince sorted → Planning by the Secretariat → Review by the Chancellery → Six Ministries executed simultaneously → Memorials and reports (30 seconds)

🐳 > **No OpenClaw? ** Run one line docker run -p 7891:7891 cft0808/edict to experience the full Kanban Demo (preset simulation data).


🤔 Why Three Provinces and Six Ministries?

Most Multi-Agent frameworks follow the following approach:

“Come on, you AIs can chat yourselves, and when you’re done, give me the results.” "

Then you get a lump of unclear processing results, impossible to replicate, audit, or intervene in.

The thinking of the Three Departments and Six Ministries is completely different — We use a system framework that has existed in China for 1,400 years:

This is not fancy metaphor; it is true decentralization and checks:

CrewAI MetaGPT AutoGen Three Departments and Six Ministries
Review Mechanism ❌ None ⚠️ Optional ⚠️ Human-in-loop ✅ ** Full-time Reviewer of the Secretariat · Can be rejected **
Real-Time Board ❌ ❌ ❌ ✅ ** Grand Council Kanban + Timeline**
Task Intervention ❌ ❌ ❌ ✅ ** Stop / Cancel / Restore **
Turnover Audit ⚠️ ⚠️ ❌ ✅ ** Complete memorial archive **
Agent Health Monitoring ❌ ❌ ❌ ✅ ** Heartbeat + Activity Detection **
Hot-Switching Model ❌ ❌ ❌ ✅ ** One-click switching of LLMs within Kanban **

This page was translated automatically. Technical details may differ from the original.

Recommended Tools

Try a different keyword or remove a filter.

Install

npx skillfish add cft0808/edict
New to MCP? Start with the mcpskill.org guide →