LibreAI / AI Agents / metamcp
// agents · metatool-ai/metamcp

metamcp

metamcp is an open-source ai agents project by metatool-ai (metatool-ai/metamcp on GitHub). It has 2,698 GitHub stars and 436 forks, is released under the MIT license, and was last updated on GitHub on 22 June 2026.

MCP aggregator, orchestrator, middleware and gateway in one Docker — compose many servers into one endpoint

★ 2.7k stars⑂ 436 forksMITupdated 2026-06
▦ visual — this project's README ships no embeddable screenshot,
so here's a placeholder. The guide below is complete regardless.
// project facts

Project facts

GitHub stars
2,698
Forks
436
License
MIT
Last updated on GitHub
22 June 2026
Category
AI Agents
Listing
Full install guide
Official site
Official site
// how it works

How it works

  1. Clarify your goal

    List the MCP servers you want behind one endpoint and which tools each audience should see.

  2. Get MetaMCP

    Clone the repo and copy example.env to .env.

  3. Install and configure

    Start it with docker compose up -d, then add your MCP server configurations in the UI.

  4. Run your first endpoint

    Create a namespace from your servers, assign it to an endpoint, and connect your client to that endpoint URL.

// features

Why people love it

Dynamic aggregation

Combines many MCP servers into one unified MCP server at runtime — no client-side config sprawl.

Namespaces

Group servers into namespaces, enable/disable servers or individual tools, and switch a namespace on an endpoint in one click.

Endpoints with auth

Publish SSE, Streamable HTTP or OpenAPI endpoints with API-key auth (header or query) or standard MCP OAuth.

Middleware

Namespace-level middleware intercepts and transforms requests/responses; a built-in example filters inactive tools to save context.

Tool overrides

Rename tools, edit titles/descriptions and attach custom annotations per namespace without losing upstream metadata.

Inspector and multi-tenancy

Built-in inspector with saved server configs, plus multi-tenancy with private/public scopes and per-user API keys.

Advertisement
// run it

Run it yourself

  1. Clone the repo

    Clone MetaMCP and enter the directory.

    git clone https://github.com/metatool-ai/metamcp.git
    cd metamcp
  2. Prepare env and start

    Copy the example env file and start the stack with Docker Compose.

    cp example.env .env
    docker compose up -d
  3. Connect a client

    Point an MCP client at your MetaMCP endpoint (example for Cursor in the README).

    {
      "mcpServers": {
        "MetaMCP": { "url": "http://localhost:12008/metamcp/<endpoint>/sse" }
      }
    }
// use cases

What you can build

  • One governed MCP endpoint for a whole team
  • Remixing tools from many servers into task-specific bundles
  • Exposing MCP tools to Open WebUI via OpenAPI
  • Inspecting and debugging MCP servers with saved configs
Advertisement
// faq

Questions, answered

Who maintains MetaMCP?

metatool-ai maintains it; docs are at docs.metamcp.com, and the README also links a community-maintained fork.

Is MetaMCP free?

Yes, it is MIT-licensed and self-hosted in one Docker Compose stack.

Does it need a paid API key?

No. It issues its own API keys for endpoint access; the MCP servers you aggregate may have their own credentials, handled via env references.

How does it compare to MCPHub?

Both are self-hosted MCP gateways. MetaMCP emphasizes dynamic namespaces, middleware and tool overrides; MCPHub emphasizes smart routing, per-user credentials and operations monitoring.

How do I get started?

Clone, cp example.env .env, docker compose up -d, add a server in the UI, create a namespace and endpoint, and connect your client.

// keep exploring

Similar tools

AI Agents

Unla

Unla MCP Gateway: turn existing APIs and MCP servers into MCP endpoints with zero code changes.

★ 2.2k starsMIT
View guide →
AI Agents

mcphub

Self-hosted MCP gateway and control plane — connect, control and operate all your MCP servers from one place

★ 2.5k starsApache-2.0
View guide →
AI Agents

fastmcp

The fast, Pythonic way to build MCP servers and clients.

★ 28.0k starsApache-2.0
View guide →
AI Agents

open-connector

An open-source auth gateway connecting 1,500+ SaaS providers to AI agents via SDK, CLI, MCP and HTTP.

★ 6.0k starsApache-2.0
View guide →
AI Agents

archestra

Archestra: the enterprise AI platform — chat UI, LLM gateway, and MCP gateway with guardrails

★ 4.4k starsNOASSERTION
View guide →
AI Agents

mcp-shrimp-task-manager

Shrimp Task Manager: an MCP server that turns natural language into structured, dependent dev tasks.

★ 2.1k starsMIT
View guide →
// alternatives

Alternatives in AI Agents

The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if metamcp is not the right fit.

  • openclaw★ 391.6k

    A personal AI that really does things — on any OS, with your data staying yours.

  • superpowers★ 294.9k

    An agentic skills framework & software development methodology that works.

  • ECC★ 275.2k

    Agent harness optimization system with skills, memory and security for Claude Code, Codex and more

  • hermes-agent★ 252.0k

    The agent that grows with you — a Python framework for adaptive AI agents.

  • deepseek-harness★ 245.5k

    DeepSeek's plugin-based agent harness where everything is a plugin.

// more in agents

More in AI Agents

Nearby AI Agents projects in the catalogue, closest in stars to metamcp — keep exploring the shelf either side of it.

  • danghuangshang★ 2.7k

    Multi-agent court inspired by ancient China: specialized agents debate and decide.

  • agent-skills★ 2.7k

    Official agent skills that teach AI coding tools to work with Supabase.

  • skillshare★ 2.7k

    Sync your AI coding setup: skills, agents, rules, MCP, hooks, everywhere.

  • yao-meta-skill★ 2.7k

    YAO: rigorous engineering, evaluation, and governance for reusable agent skills.

  • memsearch★ 2.7k

    Persistent memory layer for AI agents, backed by Markdown and Milvus.

  • LearnPrompt★ 2.7k

    Free Chinese AIGC course: prompting, agents, image, video, voice, open models.

Get metamcp

Free and open — grab the source, star the repo, and start creating.