Skip to content
Get packs

Guide · Open source

How to use Edict: install it on OpenClaw and run the dashboard

Updated · written with AI from the project's own documentation and checked against our catalogue data

Edict is an open-source multi-agent orchestration system that runs on top of OpenClaw. It arranges about 12 AI agents in roles modeled on a Chinese imperial administration, with a web dashboard for tasks, agent health, token usage and per-task history. This guide walks through the prerequisites, the install script, starting the dashboard and the main features, so you can run a first task and see how it moves through the system.

Edictcft0808/edict · ★ 17k · MIT

Quick start

  1. 1Confirm prerequisites. Install OpenClaw, make sure Python 3.10+ is available, and confirm you are on macOS or Linux.
  2. 2Clone the repository. Run git clone on the Edict repository and change into the edict folder.
  3. 3Run the installer. Make install.sh executable and run it to create the agent workspaces and register the agents.
  4. 4Add your API key. Run openclaw agents add taizi, then run ./install.sh again to sync the key to all agents.
  5. 5Start the dashboard. Run start.sh, or start the refresh loop and dashboard server separately, then open the dashboard in your browser.
  6. 6Submit and watch a task. Send an instruction and follow it on the Kanban board, where you can view its flow, pause, cancel or resume it.

Check the prerequisites before you install

Edict is built to run on top of OpenClaw, and the README lists OpenClaw as a requirement. The dashboard depends on OpenClaw for live data, so install it first.

  • OpenClaw installed (see https://openclaw.ai)
  • Python 3.10+
  • macOS or Linux (the README lists these as the prerequisites; Windows is not listed)
  • Node.js 18+ is optional, used only to build the React frontend. If it is missing, the installer skips that step.

Install Edict from the repository

Installation is clone-and-build. Clone the repository, enter the folder, then make the installer executable and run it.

  • The install script creates the agent workspaces and writes each department's SOUL.md file.
  • It registers the agents and their permission matrix in openclaw.json.
  • It links the data and scripts folders of each workspace to the project directory so data stays consistent.
  • It syncs your API key to all agents from an agent you have already configured.
  • It restarts the Gateway so the configuration takes effect.
git clone https://github.com/cft0808/edict.git cd edict chmod +x install.sh && ./install.sh

Add an API key before the first install

The README notes a first-time setup step: configure an API key by adding the taizi agent, then run the installer again so the key is synced to every agent.

  • Run openclaw agents add taizi to configure the API key.
  • Run ./install.sh again to sync the key to all agents.
openclaw agents add taizi ./install.sh

Start the dashboard and open it in a browser

You can start everything with one script, or run the data refresh loop and the dashboard server separately. The README recommends the one-command start. Once the server is running, open the dashboard address in your browser.

  • Option 1 (recommended): run start.sh.
  • Option 2: run the refresh loop and the server in separate commands.
  • The dashboard address given in the README is http://127.0.0.1:7891.
chmod +x start.sh && ./start.sh # or, separately: bash scripts/run_loop.sh & python3 dashboard/server.py

Explore the dashboard panels

The dashboard, called the 军机处 (Military Council) board in the README, has several panels. The Kanban board is the main view for tasks.

  • Kanban: tasks in status columns, filtering by department, full-text search, heartbeat badges, task details with the full flow, and pause, cancel and resume actions.
  • Monitor: task counts per status, department distribution, and live agent health cards.
  • Memorials: completed tasks archived with a five-stage timeline and a one-click copy to Markdown.
  • Models: switch the LLM used by each agent from the UI. The README says applying a change restarts the Gateway and takes about 5 seconds.
  • Skills: view installed skills per department and add new ones.
  • Officials: token usage ranking and activity statistics.

Understand how a task moves through the agents

A request from you goes to the taizi agent, which sorts it: small talk gets a direct reply, and a real instruction becomes a task. The task then passes through planning, review and dispatch before the six execution departments carry it out. The README says the review step can reject a plan and send it back for rework, and that every task passes through this review stage.

  • Taizi (太子): sorts incoming messages.
  • Zhongshu (中书省): plans and breaks work into subtasks.
  • Menxia (门下省): reviews the plan and approves it or rejects it.
  • Shangshu (尚书省): dispatches work and reports back.
  • Six departments (六部) plus the HR department (吏部): execute the work in parallel.

Run a production setup with systemd (optional)

For a server that stays running, the README includes a systemd setup and a management script.

  • Install the service file with systemd, then enable and start it.
  • Use edict.sh to start, stop, restart or check status.
sudo cp edict.service /etc/systemd/system/ sudo systemctl daemon-reload sudo systemctl enable edict sudo systemctl start edict bash edict.sh status

What people use it for

  • Running a multi-step coding or research request through planning and a review step before execution
  • Watching which agent holds a task and how long it has been idle using heartbeat badges
  • Comparing or changing the LLM used by each agent from the Models panel
  • Reviewing the full path of a finished task in the Memorials archive

Frequently asked questions

Is Edict free to use?

Edict is released under the MIT licence, as listed in the project details. The MIT licence permits use, modification and redistribution. Running the agents still requires an LLM API key and OpenClaw, and the README does not list pricing for those.

What do I need to install Edict?

You need OpenClaw installed, Python 3.10+, and macOS or Linux. Node.js 18+ is optional and only needed to build the React frontend. The README does not list Windows as supported.

Does Edict work on Windows?

The README lists macOS and Linux as the prerequisites and does not mention Windows support. Use macOS or Linux for the documented install path.

Can I run Edict without OpenClaw?

The README says OpenClaw is required and the dashboard depends on it for live data. The README also mentions a Docker image that runs the dashboard demo with preset simulated data, which does not need OpenClaw for that demo.

How is Edict different from CrewAI or AutoGen?

The README compares Edict to CrewAI, MetaGPT and AutoGen in a feature table. It highlights a built-in review step that can reject work, a real-time dashboard, the ability to pause or cancel tasks, and per-agent model switching. Those are the project's own claims, so check the table and the architecture docs before deciding.

Why does the installer ask for an API key first?

The agents need an LLM API key to run. The README says to configure a key by adding the taizi agent, then rerun install.sh so the key is copied to every other agent.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Edict suits developers who already run OpenClaw and want a multi-agent workflow with a review step, visible task history and a dashboard to pause tasks or change models. It is a poor fit for people without OpenClaw, for non-technical users who are uncomfortable with shell scripts and API keys, and for Windows users, since the README lists only macOS and Linux.

Install options & downloads