# metatool-ai/metatool-app [Health: Active]

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/metatool-ai/metatool-app  
**GitHub Stars:** 2732  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/metatool-ai-metatool-app

## Description
MetaMCP is the one unified middleware MCP server that manages your MCP connections with GUI.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "metatool-app": {
    "command": "uvx",
    "args": ["mcp-hn"]
  }
}
```

## Documentation

## What metatool-ai/metatool-app MCP server does

metatool-ai/metatool-app MCP server acts as a proxy and aggregation layer for MCP infrastructure. It stores configurations for connected MCP servers, groups them into namespaces, and emits each namespace through one or more endpoints. A namespace can contain multiple servers, and individual servers or tools can be enabled or disabled.

The system is intended for teams or developers that want to host dynamically composed MCP servers rather than configure every upstream server separately in each agent client. It also provides a graphical interface for managing connections and saved server configurations.

## How it works

A server configuration tells MetaMCP how to launch an upstream MCP server. STDIO examples use a command and argument list, such as starting a server with `uvx`. Connected servers are grouped into namespaces. Namespace settings can control which tools are available, apply request or response middleware, and change tool names, titles, descriptions, or MCP annotations.

Endpoints are assigned to namespaces. The tools from all enabled servers in the selected namespace are then presented through the endpoint as one MetaMCP server. Supported MCP transports include SSE and Streamable HTTP. The project also documents OpenAPI endpoints for clients such as Open WebUI.

Middleware operates at namespace scope and can intercept or transform MCP requests and responses. The documented built-in example filters inactive tools to reduce the tool context exposed to language models. The Inspector provides saved server configurations and can be used to examine MetaMCP endpoints.

## Setup and configuration

Docker Compose is the recommended way to run metatool-ai/metatool-app MCP server. The repository also documents development with Dev Containers in VS Code or Cursor, plus local development. Deployment details, configuration examples, and connection instructions are provided in the project documentation at docs.metamcp.com.

STDIO server environment values can be supplied directly, referenced with `${ENV_VAR_NAME}`, or passed through automatically when the expected variable name matches a variable in the MetaMCP container. Referencing container variables keeps secret values out of the server configuration and repository. For local Docker development with `pnpm run dev:docker`, variables may also need to be listed in `turbo.json` under `globalEnv`.

Endpoints support API-key authentication through a header or query parameter, as well as standard OAuth from MCP specification version 2025-06-18.

## Tools and capabilities

- Aggregate multiple MCP servers behind one endpoint.
- Organize servers into namespaces and switch the namespace assigned to an endpoint.
- Enable or disable servers and individual tools.
- Edit tool names, titles, descriptions, and namespace-specific annotations.
- Apply namespace-level middleware to MCP requests and responses.
- Inspect endpoints using saved server configurations.
- Expose endpoints through SSE, Streamable HTTP, or documented OpenAPI routes.
- Configure API-key or OAuth authentication.

## Limitations and notes

Some capabilities described by the project are future work rather than current functionality. Tool logging, error tracing, validation, scanning, and Elasticsearch-style tool selection are listed as future ideas or coming-soon use cases.

The server is infrastructure for composing and exposing MCP services; the usefulness of a namespace depends on the upstream MCP server configurations supplied to it. Clients that only support STDIO require the documented connection approach, while HTTP-capable clients can connect to the corresponding endpoint transport.

_Full upstream README: https://allmcps.com/mcp/metatool-ai-metatool-app/readme_