edict: what it is, what problem it solves & why it's gaining traction

edict: what it is, what problem it solves & why it's gaining traction

What it solves

Edict implements a multi-agent collaboration architecture inspired by the ancient Chinese "Three Departments and Six Ministries" system. It addresses the lack of transparency, auditability, and quality control found in many modern AI agent frameworks (like CrewAI or AutoGen), where agents often collaborate in a "black box" without formal oversight or the ability for humans to intervene in real-time.

How it works

The project uses a hierarchical, role-based structure to manage tasks:

  1. The Crown Prince (Taizi): Sorts incoming messages, separating casual chat from actual tasks (edicts).
  2. The Three Departments:
    • Zhongshu (Secretariat): Plans the task and breaks it down into sub-tasks.
    • Menxia (Chancellery): Acts as a mandatory quality gate, reviewing plans and either approving them or sending them back for rework (blocking).
    • Shangshu (Department of State Affairs): Dispatches tasks to the specialized ministries and aggregates the final results.
  3. The Six Ministries + HR: Specialized agents (e.g., Finance, Rites, War, Justice, Works, and Personnel) execute the specific technical or administrative tasks.

This flow is governed by a strict permission matrix and a state machine that prevents illegal task transitions. The system includes a real-time "Grand Council" (Junji Chu) dashboard for monitoring, intervening (stopping/canceling), and managing LLM configurations and skills.

Who it’s for

Developers and organizations looking for a highly structured, observable, and auditable multi-agent system where quality assurance is built into the architecture rather than being an optional plugin.

Highlights

  • Institutional Review: A dedicated agent (Menxia) specifically for auditing and rejecting subpar plans before execution.
  • Real-time Dashboard: A comprehensive Kanban board and monitor for tracking task flow, agent health, and token consumption.
  • Human Intervention: Ability to stop, cancel, or resume tasks directly from the UI.
  • Skill Management: A system to import and update specialized skills from GitHub or other URLs via UI, CLI, or API.
  • Robust Backend: Uses Redis Streams for an event bus and an Outbox Relay pattern to ensure reliable event delivery.
  • Model Flexibility: Independent LLM switching for each agent via the dashboard.

Sources