Meta MCP Search

Meta MCP Search

Provides a semantic search interface that enables discovery and routing across over 1,000 local MCP tools using natural language queries. It leverages hybrid search and embeddings to accurately match user intent with tool names and descriptions.

Category
访问服务器

README

Meta MCP Search

A single MCP server exposing search_tool that routes to 1000+ local MCP tools via semantic search.

Features

  • Semantic Search: Uses gte-small embeddings (8-bit quantized, multi-threaded) via @xenova/transformers for fast, accurate tool discovery
  • Hybrid Search: Cosine similarity on query vs tool name + description
  • MCP Protocol: Full implementation of Model Context Protocol with stdio transport
  • Dual Usage: Can be used as an MCP server or imported directly as a TypeScript module

Installation

npm install meta-mcp-search

Usage

As MCP Server (stdio) - Quick Start

The easiest way to run the MCP server is with npx:

npx meta-mcp-search

Or if installed globally:

npm install -g meta-mcp-search
meta-mcp-search

The server will:

  1. Load tools from config.json or tools-manifest.json in the current directory
  2. Build embeddings for all tools
  3. Listen on stdio for MCP requests

As Imported Module

import { MetaMcpSearch, searchToolsDirect } from 'meta-mcp-search';

// Option 1: Create instance and use directly
const metaMcp = new MetaMcpSearch({
  configPath: './tools-manifest.json'
});
await metaMcp.init();

const tools = await metaMcp.search('send a message to slack');
console.log(tools);

// Option 2: Quick search function
const tools = await searchToolsDirect('list files in google drive', {
  configPath: './config.json'
});

Direct Function Calls

import { 
  SearchEngine, 
  loadToolsFromConfig,
  initSearchEngine,
  searchTools 
} from 'meta-mcp-search';

// Load tools
const tools = await loadToolsFromConfig('./tools-manifest.json');

// Initialize search engine
await initSearchEngine(tools);

// Search
const results = await searchTools('create a github issue', 5);

Configuration

config.json Format

{
  "mcpServers": {
    "google-drive": {
      "command": "node",
      "args": ["./servers/google-drive/dist/index.js"],
      "tools": [
        {
          "name": "google_drive_list",
          "description": "List files in Google Drive",
          "inputSchema": {
            "type": "object",
            "properties": {
              "folderId": { "type": "string" }
            },
            "required": ["folderId"]
          }
        }
      ]
    }
  }
}

tools-manifest.json Format

{
  "version": "1.0.0",
  "tools": [
    {
      "name": "google_drive_list",
      "description": "List files in Google Drive",
      "inputSchema": {
        "type": "object",
        "properties": {
          "folderId": { "type": "string" }
        },
        "required": ["folderId"]
      },
      "serverKey": "google-drive"
    }
  ]
}

API Reference

MetaMcpSearch

Main class for the meta MCP search functionality.

const metaMcp = new MetaMcpSearch(options?: MetaMcpSearchOptions);
await metaMcp.init();
await metaMcp.search(query: string, limit?: number);
await metaMcp.start(); // Start MCP server

SearchEngine

Low-level search engine class.

const engine = new SearchEngine();
await engine.init(tools);
const results = await engine.search(query, limit);

loadToolsFromConfig(path?: string)

Load tools from configuration file.

const tools = await loadToolsFromConfig('./config.json');

MCP Tool: search_tool

The server exposes a single tool:

Input Schema:

{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "description": "Natural language query describing what you want to accomplish"
    },
    "limit": {
      "type": "number",
      "default": 8,
      "description": "Maximum number of results to return"
    }
  },
  "required": ["query"]
}

Output:

[
  {
    "name": "slack_send_message",
    "description": "Send a message to a Slack channel",
    "inputSchema": { ... },
    "serverKey": "slack",
    "score": 0.89
  }
]

Development

# Build
npm run build

# Development (watch mode)
npm run dev

# Clean build artifacts
npm run clean

# Run tests
npm test

# Run tests with coverage
npm run test:coverage

Publishing to npm

This package is published to npm. To publish a new version:

# 1. Make sure you're logged in to npm
npm login

# 2. Update the version in package.json
npm version patch  # or minor, or major

# 3. Build and test
npm run build
npm test

# 4. Publish
npm publish

The prepublishOnly script will automatically run clean and build before publishing.

Requirements

  • Node.js >= 18.0.0
  • npm

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选