M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
E

EmperorClaw

Community
The self-hosted operations platform for companies run with AI agents.
How to install ↓ Compare GitHub source repository ↗
Category
Other #179 of 230
Stars
★ 515 Very popular
Transport
stdio (local process)
Runtime
Node.js 20+ · Docker
Credentials
API key / credential required
License
Other / unspecified
Last commit
45FMRS · D

EmperorClaw provides a comprehensive self-hosted platform for managing AI agent operations, with runtime independence, durable task leasing, knowledge management, pipelines, and human approvals. It's ideal for organizations looking to establish a reliable, auditable record of multi-agent workflows, but requires infrastructure to run a long-running Node.js process and PostgreSQL.

Strongest · Setup experience 12/20 Weakest · Reliability 6/20

Reliability
6/20
Security and permissions
7/20
Maintenance
8/20
Documentation
12/20
Setup experience
12/20
Why each score
Reliability 6/20
Unable to verify that the server starts cleanly or completes the MCP init and tool-list handshake, as no server manifest was provided. The README declares a broad MCP API but does not provide a concrete tool list or test evidence. There are hints of CI workflows and tests, but no specifics. Deductions for lack of verifiable execution evidence and clear MCP tool surface.
Security and permissions 7/20
The codebase appears to include standard auth mechanisms (NextAuth, Argon2, JWT, bearer tokens), mentions of AES-256-GCM encryption, and path sanitization. However, no critical vulnerabilities were identified, but as a reviewer I cannot inspect the code, so permission and confirmation mechanisms are not fully transparent. Score of 7 due to invisibility of security boundaries and confirmation of dangerous operations.
Maintenance 8/20
Project is active with 275 stars, 0 open issues, conventional commits, and a clear governance model. License is FSL-1.1-Apache-2.0, which restricts competing hosted services but allows forking and modification. Update mechanisms and backup scripts exist. Deductions for lack of SECURITY.md and visibility into dependency management.
Documentation 12/20
README provides a comprehensive overview, installation instructions, architecture details, feature matrix, and configuration reference. However, it lacks specific MCP tool definitions, parameters, limits, and troubleshooting guides. The documentation does not clearly list all MCP tools and their parameters, limiting actionability. Deductions for insufficient MCP-specific coverage.
Setup experience 12/20
Installation steps are clear, with one-click Docker install, manual setup, and manual installation instructions, configuration examples, and deployment constraints explained. However, since the installer scripts cannot be verified, and the requirement of PostgreSQL and long-running process adds assumptions, the setup is plausible but potentially fragile in real environments. Clear steps but based on static review.

Static review · not runListed 2026-08-07

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesUses the networkConnects to a database

Best for

  • Digital agencies operating AI agents and needing durable operational state
  • Accounting and bookkeeping firms requiring client-isolated knowledge and audit history
  • Consultancies needing structured artifact storage and approvals
  • Ecommerce companies running AI back-office operations
  • Software companies needing multi-agent workflow coordination

Not for

  • People looking for a chatbot or prompt wrapper
  • Users needing fully serverless deployment without persistent WebSocket connections
  • Teams seeking an open-source (non-Fair Source) license
  • Users looking for a lightweight task management tool rather than an operations platform

Required permissions

  • Requires access to a PostgreSQL database for state storage
  • Requires ability to maintain persistent WebSocket connections
  • May require local filesystem or external storage backend (e.g., Bunny CDN) for artifacts
  • Requires secrets in environment variables (e.g., NEXTAUTH_SECRET, EMPEROR_CLAW_MASTER_KEY)

Risks and side effects

  • Self-hosting: risk of data loss if not properly configured; regular backups needed
  • Not open source (Fair Source) — restriction on offering it as a competing commercial product/service
  • Requires Node.js 20+ and PostgreSQL 16 as prerequisites
  • Not compatible with serverless platforms; requires a long-running process like a VPS or dedicated server

Setup

Before you start

Runtime：Node.js 20+ · Docker

POSTGRES_CONNECTION_STRING requiredsecret PostgreSQL 16 connection string, required.
NEXTAUTH_SECRET requiredsecret Session encryption secret, generate with openssl rand -base64 32, required.
EMPEROR_CLAW_MASTER_KEY optionalsecret AES-256-GCM master key encrypting integration secrets at rest, recommended; generate with openssl rand -hex 32.
Other optional settings (7)
DEPLOYMENT_MODE optional Deployment mode: self-hosted (default) or cloud.
STORAGE_BACKEND optional Storage backend: local (default) or bunny.
DRIVE_SYNC_ENABLED optional Enables local shared-volume Drive discovery; requires the optional drive Compose profile.
DRIVE_RCLONE_REMOTE optional rclone remote name, default gdrive.
DRIVE_ROOT optional Destination folder in Drive, default Emperor-Claw.
SMTP_HOST optional SMTP server for invitations and password reset; disabled if unset.
EMPEROR_PLATFORM_ADMIN_EMAILS optional Comma-separated emails granted /ops platform admin access.
  1. Run the one-command install script with curl -fsSL https://raw.githubusercontent.com/emperorclaw/emperorclaw/main/install.sh | bash, or the PowerShell script for Windows.
  2. Open http://localhost:3000 and create your admin account. Alternatively, manually clone the repo, copy .env.example to .env, fill in the required environment variables, and run npm install, npm run db:migrate, npm run build, and npm start.

Check that it works

After install, open http://localhost:3000 and create the admin account, generate an MCP token at Settings → Tokens, configure your agent, and confirm it registers, heartbeats, and claims tasks in the dashboard.

Troubleshooting

  1. Ensure Node.js version is >= 20
  2. Check that PostgreSQL connection string is correct
  3. Confirm migrations have been run with `npm run db:migrate`
  4. Verify WebSocket connections are not blocked by firewall
  5. Check that the watchdog process is running (ensured via Postgres advisory lock)

Things to try

Once connected, you can ask your AI assistant things like:

  • Register my agent through the MCP API and claim the next available task
  • Show the live status and heartbeats of all registered agents
  • Create a new task in the project and assign it to an agent
  • Search the Company Brain for the SOPs scoped to this customer

Use cases

Create a durable, auditable operating record for AI agent workflows
Coordinate projects and tasks across multiple clients with leases and approval gates
Provide agents with context before they act via scoped knowledge (Company Brain)
Track task lifecycles through task leases, retries, and dead-letter queues
Visualize and monitor AI workflows through pipeline registration and run reports

Overview

EmperorClaw is a self-hosted, runtime-neutral operations platform for coordinating agents, customers, projects, knowledge, artifacts, approvals, and incidents. It provides durable state around agent execution, serving as a single system of record for the entire workforce. It does not execute agent logic or call LLMs; instead, it coordinates, records, and surfaces everything around that execution. It uses PostgreSQL for state, WebSocket for realtime features, and includes a watchdog process for lease expiry and dead letters.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 60aec6f0306a Data synced 2026-10-19 Read the FMRS scoring method