当皇上 (danghuangshang)
Open-source multi-agent collaboration system that models Chinese imperial governance as a team of specialized AI agents running on OpenClaw.
At a Glance
Self-hosted MIT-licensed multi-agent system; users supply their own Anthropic API key and server (external costs).
Engagement
Available On
Alternatives
Listed Oct 2026
About 当皇上 (danghuangshang)
danghuangshang (当皇上) is an MIT-licensed multi-agent collaboration system published by wanikua. It maps Chinese imperial governance structures, such as the Ming cabinet and Tang three-departments model, onto specialized AI agents that you command from Discord, Feishu or a web UI. The repository's latest listed release is v3.0 and the README references v3.5.3.
What It Is
The project is a ready-to-deploy multi-agent team built on the OpenClaw framework. You act as the emperor and issue instructions by mentioning an agent; a dispatcher agent (司礼监) receives the task, a cabinet agent refines the prompt and produces a plan, and ministry agents carry it out. Ministries cover engineering, finance, marketing, DevOps, legal, HR and project management, and a censorate agent reviews code pushed to GitHub.
Governance Modes
The installer offers three regimes: Ming cabinet (default, fast iteration), Tang three-departments (draft, review with veto, execute) and a modern corporate model with CEO/CTO/CFO roles in English. A script switches between regimes after installation. A separate Hanlin Academy group of five agents can collaborate on writing novels.
Capabilities and Setup
- Each agent has its own workspace and memory, and different agents can use different models such as Anthropic, OpenAI, Gemini or DeepSeek.
- The README lists 60+ built-in skills, including GitHub, Notion, browser automation, cron jobs, TTS and shell control.
- Docker sandboxing isolates agent code execution.
- A React/TypeScript web dashboard shows court status, sessions, cron jobs and token usage.
- Installation is via a shell script (local clone or remote one-liner), a lite script for existing OpenClaw users, a Hermes Agent runtime variant, or a Docker image. You supply an LLM API key and Discord bot tokens.
Related Projects
The site describes Tiangong (天工开物) as a separate engine layer with self-evolving prompts and OpenViking memory, and Become CEO as an English enterprise version. It also announces a hosted Agent Team as a Service as coming soon.
Community Discussions
Be the first to start a conversation about 当皇上 (danghuangshang)
Share your experience with 当皇上 (danghuangshang), ask questions, or help others learn from your insights.
Pricing
Open Source (MIT)
Self-hosted MIT-licensed multi-agent system; users supply their own Anthropic API key and server (external costs).
- MIT license, commercial use unrestricted
- Self-hosted; data stays on your own server
- External costs: Anthropic API key approx. $5-20/month and a server (BYOK/hosting)
- Supports Discord, Feishu, WhatsApp, Web
Capabilities
Key Features
- Multi-agent collaboration with dispatcher, cabinet and ministry agents
- Three selectable governance regimes (Ming cabinet, Tang three-departments, modern corporate)
- Cabinet prompt enhancement before task dispatch
- Automatic code review on GitHub push via censorate agent
- Per-agent independent memory and workspace
- 60+ built-in skills
- Cron scheduled tasks
- Docker sandbox isolation
- Web dashboard for sessions, cron and token stats
- Memory backup scripts
- Multiple models per agent
- Zero-code agent configuration via config file