emperorclaw/emperorclaw

EmperorClaw – a self‑hosted control‑center for AI‑agent workforces

What it is – A single Node.js/Next.js application (Docker‑installable) that sits between people and any LLM‑based agents you run. It provides a web dashboard and an API so agents can register, claim tasks, exchange messages, and store results while humans can see a kanban‑style board, a knowledge base, file storage, chat, and an audit trail.

Why it exists – When you have more than one autonomous agent, you quickly lose visibility: you don’t know which agent is doing what, whether a task failed, where outputs are stored, or which rules applied. EmperorClaw records all of that state in PostgreSQL and surfaces it in a UI, giving you the same organisational tools you use for human teams.

Core capabilities

Feature	What it gives you
Task board	Kanban view, owners, statuses, approval gates
Customer directory	Separate context, files, and knowledge per client
Company Brain	Wiki‑style knowledge base scoped to company, project, client or agent
Secure storage	Hierarchical file store (local by default, optional Bunny CDN or S3‑compatible back‑ends)
Team chat	Persistent, routable conversations between humans and agents
Pipelines	Repeatable, visual workflows with approval steps and full audit logs
Agent lifecycle	Registration, heart‑beats, automatic retries, incident flagging
Access control	Invite‑only sign‑up, role‑based permissions, per‑company API tokens
Self‑hosted	Runs on any VPS/VM with Docker, PostgreSQL 16, Node ≥20; no cloud lock‑in

Who should consider it – Agencies, bookkeeping firms, consultancies, e‑commerce or software companies, recruiting firms, and any small business that wants an AI‑back‑office that is auditable and centrally managed.

Getting started – The repo ships a one‑line Docker installer:

# mac/linux
curl -fsSL https://raw.githubusercontent.com/emperorclaw/emperorclaw/main/install.sh | bash
# windows (PowerShell)
irm https://raw.githubusercontent.com/emperorclaw/emperorclaw/main/install.ps1 | iex

The script creates a PostgreSQL container, runs migrations, generates secrets, and starts the web app on http://localhost:3000. After signing up you can:

1. Hire an agent (choose a role, give an LLM API key).
2. Verify the agent is online via a simple “Hello” message.
3. Create a project, add a task, assign it, and watch the progress on the board.

Extensibility – EmperorClaw talks to agents through a generic MCP API. It ships adapters for:

• OpenClaw (native bridge)
• Hermes (browsing/scraping runtime)
• Any MCP‑compatible runtime you build.

Security & privacy – Passwords hashed with Argon2, JWT sessions, SHA‑256‑hashed API tokens, master‑key encryption for stored LLM credentials, path‑traversal‑protected storage, rate‑limiting, and a “last‑admin guard” to prevent lock‑out.

License – Fair‑Source License (FSL‑1.1) that allows self‑hosting, modification, and commercial use, but forbids reselling the product as a competing hosted service. Two years after a release the code automatically re‑licenses to Apache 2.0.

Where to learn more –

• Docs: https://github.com/emperorclaw/emperorclaw (includes installation guide, first‑agent walkthrough, and architecture details)
• Discussions: https://github.com/emperorclaw/emperorclaw/discussions
• Hosted demo on Render: https://render.com/deploy?repo=https://github.com/emperorclaw/emperorclaw

---

Bottom line – EmperorClaw is a genuine, production‑ready platform for orchestrating, tracking, and auditing multiple AI agents, giving enterprises the same organisational scaffolding they use for human workers while keeping all data under their own control.