AI Frameworks · cft0808

edict

Edict is a Python-based multi-agent orchestration system inspired by ancient imperial governance structures, featuring 12 specialized AI agents, a real-time Kanban dashboard, and mandatory audit/review layers for agent output validation and traceability.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/cft0808/edict

16.2k

GitHub stars

1.7k

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	cft0808/edict
Owner	cft0808
Primary language	Python
License	MIT — OSI-approved
Stars	16.2k
Forks	1.7k
Open issues	38
Latest release	Unknown
Last updated	2026-07-06
Source	https://github.com/cft0808/edict

What edict is

Built on OpenClaw, Edict implements a hierarchical agent architecture (taizi→zhongshu→menxia→shangshu→liubu) with state-machine-enforced task routing, audit trails, and a React 18 frontend dashboard that enables live LLM switching, skill management, and task intervention without external backend dependencies.

Quickstart

Get the edict source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/cft0808/edict.git2cd edict3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Complex multi-stage AI workflows requiring auditability

Tasks requiring approval gates (e.g., compliance-sensitive content generation, incident response orchestration) benefit from the mandatory menxia (審議) review layer that blocks non-compliant outputs and enforces rework cycles.

Distributed team collaboration with clear role boundaries

Organizations needing explicit permission matrices, capability segregation, and real-time visibility into agent state can use the out-of-box Kanban + monitoring dashboard to manage 12 specialized agents across provinces (departments).

Production AI agent deployments needing observability and intervention

The dashboard provides pause/cancel/resume controls, per-agent model hot-swapping, session monitoring, and complete transaction logs (奏折) — critical for mission-critical automation where black-box execution is unacceptable.

Implementation considerations

OpenClaw must be pre-installed and configured with valid API keys; edict-provided install.sh attempts sync across all 12 agent workspaces but can fail silently if OpenClaw setup is incomplete.
Agent-to-agent communication requires explicit visibility settings (sessions.visibility all); default OpenClaw configs may prevent message delivery between menxia (审議) and shangshu (尚書) agents.
State machine enforcement via kanban_update.py rejects invalid transitions; ops teams must understand the 9-state model (pending→draft→review→approved→executing→…→archived) or task progression will stall.
React frontend build (Node.js 18+) is optional but recommended; docker-compose provides pre-built frontend, but source deployments require npm install in dashboard/ directory.
Performance: Real-time monitoring loop (scripts/run_loop.sh) polls OpenClaw at fixed intervals; latency scales with agent count and task volume. No horizontal scaling documented.

When to avoid it — and what to weigh

Need simple, lightweight agent coordination — If your task is straightforward (e.g., single LLM call orchestration), the 12-agent hierarchical model adds unnecessary operational overhead. CrewAI or simpler frameworks are more appropriate.
OpenClaw is not available or not your platform choice — Edict is tightly coupled to OpenClaw (required dependency per README). If your org uses different LLM platforms or agent runtimes, this is a blocking constraint.
Need immediate, out-of-the-box plug-and-play setup — Installation requires Workspace creation, SOUL.md configuration, permission matrix registration, API Key synchronization across 12 agents, and Node.js for frontend builds. Plan 1–2 hours of setup and 2–4 weeks of refinement for production.
Require established, battle-tested multi-agent framework — Project created Feb 2026, first stable release unknown (latestRelease: n/a), with 38 open issues. Not recommended for risk-averse deployments; consider AutoGen or MetaGPT if maturity/stability is critical.

License & commercial use

MIT License. Permissive: permits commercial use, modification, and redistribution with attribution and no warranty clause.

MIT is a permissive OSI license that explicitly allows commercial use, including building closed-source products or SaaS on top of Edict. No runtime royalties or commercial restrictions are imposed. However, attribution in documentation and code headers is required. Review your legal requirements for bundled dependencies (React, backend stdlib).

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	Medium

Security considerations

No explicit security audit or threat model is documented. Skill system allows arbitrary Python execution on agents; no mention of sandboxing, allowlisting, or audit logging for skill invocation. Agent-to-agent message passing relies on OpenClaw's access control (not reviewed here). Permission matrix is human-configured (kanban_update.py enforces state rules, not data access control). For production use, conduct a security review of skill handlers and agent communication paths.

Alternatives to consider

CrewAI

Simpler, more mature multi-agent framework with built-in role definitions and tool/skill management. No audit layer, but significantly lower operational overhead and larger ecosystem. Better for teams without specialized audit requirements.

AutoGen (Microsoft)

Established multi-agent orchestration with human-in-the-loop and conversation management. Supports multiple LLM backends and has wider adoption. Less intrusive state machine; more flexible for custom workflows.

MetaGPT

Task decomposition and role-based multi-agent system with Software Engineering framework. Stronger for coding/development tasks. Less emphasis on audit trails, but comparable feature richness in agent specialization.

Software development agency

Build on edict with DEV.co software developers

Start with the Docker demo (one command), then evaluate whether Edict's hierarchical governance model and audit-first design fit your production AI workflows. Requires OpenClaw.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

agentops agenticSeek crewAI

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development Enterprise Software Development Custom Software Development Talk to DEV.co

edict FAQ

Do I need OpenClaw to run Edict?

Yes. OpenClaw is a hard dependency (explicitly stated in README badge and installation docs). Without it, Edict agents cannot execute. If you don't have OpenClaw, use the demo Docker image (cft0808/sansheng-demo) to explore the dashboard with pre-loaded data.

Can I use Edict without the dashboard?

Unknown. The system is designed around the Kanban UI for task intervention and monitoring. The backend (OpenClaw agents) can operate independently, but observability and control are severely diminished. Not recommended.

What happens if the menxia (審議) agent rejects a task?

The zhongshu (中書省) agent is expected to revise the task plan and resubmit. The framework enforces a rework loop; tasks cannot proceed to execution until menxia approves. Details of retry logic and max-rejection thresholds are not clearly documented.

How do I add a custom skill to the liubu (六部) agents?

Skill management is exposed in the dashboard UI, and installation is mentioned as 'adding new skills.' The actual implementation (where to place skill code, format, registration) is not detailed in the provided excerpts. Refer to the docs/task-dispatch-architecture.md or CONTRIBUTING.md for guidance.

Custom software development services

Adopting edict is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate ai frameworks software in production.

Explore Multi-Agent Orchestration with Audit & Control

Start with the Docker demo (one command), then evaluate whether Edict's hierarchical governance model and audit-first design fit your production AI workflows. Requires OpenClaw.

Talk to DEV.co Schedule a Consultation