# MetaMCP — Aggregate Many MCP Servers as One

> MetaMCP aggregates multiple MCP servers into namespaces and exposes a single endpoint with tool overrides, middleware, and an inspector for debugging.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

1. Clone + create env file:
   ```bash
   git clone https://github.com/metatool-ai/metamcp.git
   cd metamcp
   cp example.env .env
   ```
2. Start with Docker Compose:
   ```bash
   docker compose up -d
   ```
3. Add servers to a namespace, then expose a MetaMCP endpoint for clients.


## Intro

MetaMCP aggregates multiple MCP servers into namespaces and exposes a single endpoint with tool overrides, middleware, and an inspector for debugging.

- **Best for:** teams building an MCP ‘hub’ with per-namespace policies, overrides, and debugging tools
- **Works with:** Docker + docker compose, Postgres (bundled), MCP servers as upstreams
- **Setup time:** 20–45 minutes

## Practical Notes

- Data point: the recommended path is Docker Compose + a bundled Postgres volume.
- Quant: keep namespaces small (5–15 tools) so tool context stays usable for LLMs.

## Pattern: enforce policy at the namespace boundary

When you combine many upstream servers, the namespace becomes your policy layer:
- filter or hide tools,
- normalize tool names/descriptions,
- and add logging.

## Pattern: create an “inspection loop”

Run your client against the MetaMCP endpoint, then use the inspector to replay calls with saved configs. This turns tool integration from guesswork into a debuggable loop.

### FAQ

**Q: Is MetaMCP a single binary?**
A: It’s designed to run via Docker Compose (recommended) with supporting services.

**Q: Why do tool overrides matter?**
A: Small metadata tweaks can drastically improve tool selection quality.

**Q: How do I avoid tool explosion?**
A: Use namespaces + filters to keep only the tools users actually need.

## Source & Thanks

> Source: https://github.com/metatool-ai/metamcp
> License: MIT
> GitHub stars: 2,315 · forks: 339

---

<!-- ZH -->

## 快速使用

1. 克隆并准备环境变量：
   ```bash
   git clone https://github.com/metatool-ai/metamcp.git
   cd metamcp
   cp example.env .env
   ```
2. 用 Docker Compose 启动：
   ```bash
   docker compose up -d
   ```
3. 把多个 server 加进 namespace，再对外发布一个 MetaMCP endpoint。


## 简介

MetaMCP 可以把多个 MCP server 聚合到 namespace，再对外输出一个统一端点；支持 tool 覆盖、middleware，并提供可保存配置的调试 inspector 便于复盘。

- **适合谁：** 想做 MCP ‘中枢’、并按 namespace 管理策略/覆盖/调试能力的团队
- **可搭配：** Docker + docker compose、内置 Postgres、上游 MCP servers
- **准备时间：** 20–45 分钟

## 实战建议

- 数据点：推荐路径是 Docker Compose，并带一个 Postgres volume。
- 量化建议：让每个 namespace 的工具数量保持在 5–15 个，避免上下文膨胀。

## 常用打法：在 namespace 边界做策略

聚合多个上游时，namespace 就是策略层：
- 过滤/隐藏工具，
- 统一工具命名与描述，
- 增加日志与可观测性。

## 常用打法：建立“可复现的调试闭环”

让客户端调用 MetaMCP endpoint，再用 inspector 复盘调用并保存配置，把工具接入从“猜”变成可调试的流程。

### FAQ

**MetaMCP 是单个二进制吗？**
答：推荐用 Docker Compose 运行，并带配套服务。

**为什么要做 tool 覆盖？**
答：少量元数据调整就能显著改善模型选工具的准确率。

**怎么避免工具爆炸？**
答：用 namespace + filter 只暴露真正需要的工具。

## 来源与感谢

> Source: https://github.com/metatool-ai/metamcp
> License: MIT
> GitHub stars: 2,315 · forks: 339


---
Source: https://tokrepo.com/en/workflows/metamcp-aggregate-many-mcp-servers-as-one
Author: MCP Hub