Repolode
Submit
cft0808

cft0808/edict

A multi-agent orchestration system for OpenClaw that models AI agent collaboration on the Tang-dynasty Three Provinces and Six Ministries structure, with a review gate and live kanban dashboard.

View on GitHubWebsite
Stars
17,009
Forks
1,765
Open issues
15
Last commit
24 days ago

What is it?

Edict is a multi-agent orchestration system built on OpenClaw. It arranges about a dozen AI agents into a hierarchy modeled on the Three Provinces and Six Ministries: a dispatcher receives requests, a planning agent drafts subtasks, a review agent can reject plans and send them back for rework, and an executing agent group carries out the approved work and reports results back.

The project's main difference from typical agent frameworks is that every task passes a mandatory review step before execution, and the permitted message routes between agents are defined in an explicit permission matrix. Task state transitions are validated by a state machine, and state changes are written to an audit log so a full task history can be traced.

A React-based dashboard shows tasks on a kanban board, displays agent health and heartbeats, supports pausing, cancelling or resuming tasks, and lets users switch the LLM used by each agent. The backend uses only the Python standard library for the dashboard server, and the project also ships with a Docker demo image that runs with simulated data.

Key features

  • 01Mandatory review stage where a dedicated agent can reject and send back plans
  • 02Permission matrix that restricts which agents can message which others
  • 03State machine validation and audit logs for every task status change
  • 04Kanban dashboard with task detail timelines, heartbeat badges and pause/cancel/resume controls
  • 05Per-agent LLM model switching and skill management from the dashboard
  • 06Prebuilt task templates and a Docker demo that runs without OpenClaw

Use cases

  • Running a multi-step software task (design, implementation, testing, docs) through separate planning, review and execution agents
  • Auditing and reviewing how an AI agent workflow reached its result
  • Experimenting with role-based agent orchestration on top of OpenClaw
  • Monitoring agent health, token usage and task progress in one dashboard

Topics

#multi-agent#ai-orchestration#workflow-automation#kanban#dashboard#llm#python#openclaw#ai-agents#autonomous-agents

FAQ

What is edict?
A multi-agent orchestration system for OpenClaw that models AI agent collaboration on the Tang-dynasty Three Provinces and Six Ministries structure, with a review gate and live kanban dashboard. Edict is a multi-agent orchestration system built on OpenClaw. It arranges about a dozen AI agents into a hierarchy modeled on the Three Provinces and Six Ministries: a dispatcher receives requests, a planning agent drafts subtasks, a review agent can reject plans and send them back for rework, and an executing agent group carries out the approved work and reports results back.
Is edict free to use? What license does it have?
edict is open source under the MIT license. Check the license terms before using it in commercial products.
Is edict still maintained?
Actively maintained. The last commit to edict was on Sep 28, 2026.
What can I use edict for?
Running a multi-step software task (design, implementation, testing, docs) through separate planning, review and execution agents Auditing and reviewing how an AI agent workflow reached its result Experimenting with role-based agent orchestration on top of OpenClaw Monitoring agent health, token usage and task progress in one dashboard
Where can I get edict?
The source code and installation instructions are on GitHub: https://github.com/cft0808/edict

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects