# MCPfinder MCP server

> Allows an AI client to discover, inspect, and install MCP servers from the Official MCP Registry and Glama.

- Repository: https://github.com/mcpfinder/mcpfinder
- Page: https://hostedmcp.org/servers/mcpfinder/server
- Transport: stdio
- Stars: 14
- License: AGPL-3.0
- Latest version: 2.0.4

## Install

### Claude Code

```sh
claude mcp add server -- npx -y @mcpfinder/server
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "server": {
      "command": "npx",
      "args": [
        "-y",
        "@mcpfinder/server"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "server": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@mcpfinder/server"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"server","type":"stdio","command":"npx","args":["-y","@mcpfinder/server"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "server": {
      "command": "npx",
      "args": [
        "-y",
        "@mcpfinder/server"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add server -- npx -y @mcpfinder/server
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.server]
command = "npx"
args = ["-y", "@mcpfinder/server"]
```

### Gemini CLI

```sh
gemini mcp add -s user server npx -- -y @mcpfinder/server
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "server": {
      "command": "npx",
      "args": [
        "-y",
        "@mcpfinder/server"
      ]
    }
  }
}
```

## Tools (4)

- `search_mcp_servers`: Call this first whenever the user needs a capability you do not already have. Use it when the user mentions an external service, a database, a filesystem, a SaaS tool, or asks "can you connect to X?". Examples: "connect to postgres" -> query "postgres"; "read my Slack" -> query "slack"; "browse local files" -> query "filesystem". If results are weak, retry with a broader term. After choosing a candidate, call get_server_details before recommending or installing it.
- `get_server_details`: Always call this before recommending or installing a server. It returns the metadata you need to judge safety and fit: install method, transport, local environment variables, remote HTTP headers and whether header metadata is known, source count, trust signals, warnings, and any discovered tools/capabilities. Use it to warn the user about secrets, stale projects, or weak metadata before calling get_install_config.
- `get_install_config`: Use this after get_server_details to generate a client-specific JSON config snippet (claude-desktop, cursor, claude-code, cline, windsurf, or vscode for VS Code's native .vscode/mcp.json). Returns paths, local env vars or remote HTTP headers (required and optional), remoteHeadersKnown, restart guidance, and safety flags. Unknown header metadata requires manual verification and sets requires_user_secrets to null.
- `browse_categories`: Single-call category browser for broad discovery. Call with no `category` to list all categories with counts; call with a `category` to get the highest-signal servers in that category. Prefer search_mcp_servers when the user names a concrete technology like Slack, Postgres, GitHub, or Notion; use this when discovery is domain-driven (database, filesystem, communication, security, ai, etc.).

## FAQ

### How do I install MCPfinder?

You can install it by adding a specific JSON snippet to your client's MCP configuration file, such as Claude Desktop, Cursor, or Claude Code.

### What tools does MCPfinder expose to my AI?

It provides four tools: searching MCP servers, getting server details, retrieving installation configurations, and browsing categories.

### Does the server block startup during the initial download?

No, the server answers the initialize request immediately and pulls the snapshot in the background to prevent handshake timeouts.
