# particlefuture/MCPDiscovery [Health: Active]

**Category:** 🔗 Aggregators  
**Repository:** https://github.com/particlefuture/1mcpserver  
**GitHub Stars:** 51  
**npm Downloads (last month):** 59  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/particlefuture-mcpdiscovery

## Description
MCP of MCPs. Automatic discovery and configure MCP servers on your local machine.

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

```json
"mcpServers": {
  "mcpdiscovery": {
    "command": "npx",
    "args": ["-y","@1mcpserver/1mcpserver"]
  }
}
```

## Documentation

## What particlefuture/MCPDiscovery MCP server does

particlefuture/MCPDiscovery MCP server acts as an MCP entry point for discovering other MCP servers. It can search a catalog of servers, identify candidates for a requested capability, and help prepare the configuration needed to use them. The catalog draws from several public MCP server collections, including awesome-mcp-servers repositories, Metorial MCP containers, and the official Model Context Protocol server repository.

The project supports two discovery styles. Quick Search is intended for a clearly stated requirement, such as finding an MCP server for payments. Deep Search handles broader goals by dividing them into components and looking for servers that cover each part. When no suitable server is found for a component, the workflow can ask whether to skip it, decompose it further, or implement it directly.

## How it works

A deep-search workflow has three stages. Planning identifies candidate servers, required keys, and configuration changes. Testing verifies server templates through `test_server_template_code`. Acting then runs the workflow with the servers that have been configured. The README also presents `deep_search` as a tool an agent can call after describing the desired outcome.

The server can be used remotely at the hosted MCP endpoint or run locally. Remote use lets an MCP client connect with a URL and HTTP headers. Local deployments can expose streamable HTTP through Docker or use standard input/output, including the source-based `--local` mode. The local approach is useful when the client only supports STDIO or when keeping the service on the local machine is preferred.

## Setup and configuration

The simplest option is to add the hosted URL, `https://mcp.1mcpserver.com/mcp/`, to an MCP client configuration. The documented examples cover Claude Desktop, Cursor, Gemini CLI, and Codex. The remote entry includes `Accept: text/event-stream` and `Cache-Control: no-cache` headers.

For local use, Docker runs the published image on port 8080, while npx launches `@1mcpserver/1mcpserver`. A source installation clones the repository, runs `uv sync`, and starts `uv run server.py --local`. The client can then invoke the process with the configured path to `uv` and repository directory.

If discovered workflows need filesystem access, configure a separate MCP filesystem server and restrict it to the directory the agent should access.

## Tools and capabilities

- Find MCP servers for explicit capabilities.
- Break larger goals into searchable components.
- Identify server keys and configuration changes during planning.
- Test server template code before use.
- Execute workflows through configured servers.
- Run through remote HTTP, Docker, npx, uvx, or local source deployment.

## Limitations and notes

Discovery and execution depend on the candidate servers selected by the workflow and any credentials or extra setup those servers require. STDIO operation through Docker may introduce connection delays. If a virtual environment reports an import error after installation, the troubleshooting guidance recommends recreating the environment.

The project lists server-to-server messaging, improved database fields, and other changes as future work rather than current capabilities. particlefuture/MCPDiscovery MCP server therefore fits best when an agent needs a discovery and configuration layer; it should not be treated as a replacement for every underlying MCP server it finds.

_Full upstream README: https://allmcps.com/mcp/particlefuture-mcpdiscovery/readme_