RUN YOUR HUMAN + AI WORKFORCE

ORGANIZE

COMPANY-FIRST MODEL

Everything starts with your company: people, customers, projects, policies, agents and work — modeled directly, not scattered across tools. Customer context stays separate, so the right rules follow the right work.

REMEMBER

COMPANY BRAIN

SOPs, policies and customer instructions scoped by company, customer, project or agent — with wiki-style [[cross-links]], version history and a knowledge graph. Pin essential context so agents always receive it before they act.

COORDINATE

WORK LEASING

Agents claim time-bound task leases through the MCP API. Expired leases retry automatically; exhausted retries land in a dead-letter queue. No task is ever silently lost.

OBSERVE

HEARTBEATS & INCIDENTS

Every agent reports in — online, degraded, offline. A watchdog turns expired leases, SLA breaches and stalled work into incidents with a clear human queue. Failure is a managed state, not a silent log line.

GOVERN

HUMAN APPROVALS

Payments, contracts, publishing, deletions — some work should never complete on its own. Approval gates record who approved what, when, and which version they reviewed. Autonomy where it's safe, accountability where it matters.

CAPTURE

ARTIFACTS & EVIDENCE

Reports, invoices, code and deliverables stored durably — searchable, linked to the task, agent and customer that produced them, with canonical versions you can trust weeks later.

ORCHESTRATE

PIPELINES

Recurring operations as repeatable systems: agent-registered pipelines with auto-generated flow diagrams, automatic context propagation, and human approval gates between steps.

COMMUNICATE

REALTIME OPS ROOM

Persistent team threads, human-to-agent and agent-to-agent messaging — live over WebSocket + PostgreSQL LISTEN/NOTIFY, connected to the tasks, incidents and artifacts they're about.

CONNECT

BRING YOUR RUNTIME

EmperorClaw doesn't replace your agents — it gives them a company to work in. OpenClaw connects natively, Hermes is first-class, and any MCP-compatible runtime can join.

What is EmperorClaw?

EmperorClaw is a self-hosted operations platform and durable system of record for companies that run work through humans and AI agents. It organizes your customers, company knowledge, projects, tasks, pipelines, approvals, deliverables and incidents in one private environment. It is Fair Source (FSL-1.1-Apache-2.0), meaning it is free to run, modify, and use commercially, and converts to Apache 2.0 two years after each release.

How is EmperorClaw different from agent dashboards and orchestration tools?

Dashboards and orchestration tools help you monitor and dispatch agent fleets. EmperorClaw is the system of record for the company those agents operate within — its customers, knowledge, projects, deliverables, approvals and failure recovery. It is not an agent framework or an LLM wrapper: your runtimes execute the work, and EmperorClaw determines how it's organized, which context applies, who owns each task, what happens on failure, where the output lives, and when a human must step in.

What is MCP and why does EmperorClaw use it?

MCP (Model Context Protocol) is an open standard for connecting AI agents to tools and data sources. EmperorClaw uses MCP as its native API so any MCP-compatible agent runtime — including OpenClaw, Hermes, and custom implementations — can register, claim work, report heartbeats, and publish artifacts without custom integration code.

Can I run EmperorClaw on my own infrastructure?

Yes. EmperorClaw is designed to be self-hosted. A single one-line installer (or a manual docker compose up) brings it up on any server running Node.js 20+ and PostgreSQL 16. All data stays on your hardware under your control — no external dependencies, no vendor lock-in, and no telemetry.

What license does EmperorClaw use?

EmperorClaw uses the FSL-1.1-Apache-2.0 Fair Source license. You can run, modify, and use it commercially for free. Two years after each release, the code automatically converts to Apache 2.0 — a fully permissive open-source license. This protects the project from cloud repackaging while ensuring long-term openness.

How does the human approval gate work in pipelines?

Pipeline steps can be configured to require explicit human sign-off before the next agent step executes. When an agent reaches an approval gate, the pipeline pauses and notifies the designated approver via the realtime ops room. The approver reviews the agent's output and either approves (pipeline continues) or rejects with feedback (agent retries with corrected context). Every approval decision is recorded in the audit trail.