@agent-infra/mcp-client

@agent-infra/mcp-client

✨ A unified MCP Client implemented in TypeScript, supporting four major transports out of the box: In-memory, Stdio, SSE (Server-Sent Events), and Streamable HTTP.

🚀 Features

• 🟦 Written in TypeScript: Type-safe, modern, and easy to integrate.
• 🔌 Multi-Transport Support: Out-of-the-box support for four major transports:
  • 🧠 In-memory: For fast, local tool integration.
  • 🖥️ Stdio: Communicate with tools via standard input/output, perfect for process-based tools.
  • 🔄 SSE (Server-Sent Events): Real-time, event-driven communication over HTTP.
  • 🌐 Streamable HTTP: Efficient, stream-based HTTP communication for scalable remote tools.
• 🛠️ Unified API: Interact with all transports using a single, consistent interface.
• 🧩 Highly Extensible: Easily add custom transports or tools as needed.

⚡ Quick Start

import { MCPClient } from '@agent-infra/mcp-client';

// type: module project usage
import { createServer as createFileSystemServer } from '@agent-infra/mcp-server-filesystem';
// commonjs project usage
// const { createServer as createFileSystemServer } = await import('@agent-infra/mcp-server-filesystem')

const mcpClient = new MCPClient([
  // In-memory
  {
    type: 'builtin',
    name: 'FileSystem',
    description: 'filesystem tool',
    mcpServer: createFileSystemServer({
      allowedDirectories: [omegaDir],
    }),
  },
  // stdio
  {
    type: 'stdio',
    name: 'FileSystem-Stdio',
    description: 'filesystem tool',
    command: 'npx',
    args: [
      '-y',
      '@agent-infra/mcp-server-filesystem'
    ]
  },
  // sse
  {
    type: 'sse',
    name: 'FileSystem-sse',
    description: 'filesystem tool',
    url: 'http://localhost:8889/sse'
  },
  // streamable-http
  {
    type: 'sse',
    name: 'FileSystem-http',
    description: 'filesystem tool',
    url: 'http://localhost:8889/mcp'
  }
]);


await mcpClient.listTools();
const result = await mcpClient.callTool({
  client: 'FileSystem-sse',
  name: 'list_directory',
  arguments: {
    path: '~/your_computer'
  },
});

🙏 Credits

Thanks to:

• kangfenmao for creating a great AI chatbot product Cherry Studio from which we draw a lot of inspiration for browser detection functionality.
• The @modelcontextprotocol/sdk project which helps us develop and use the agent tools better.