mcp client typescript

# MCP Docs Server

This project implements a Model Context Protocol (MCP) server in Node.js that provides a tool to search and fetch documentation snippets for popular AI/ML libraries.

---

## 🚀 Features

✅ MCP-compliant server (using `@modelcontextprotocol/sdk`)  
✅ Tool: **get_docs**  
✅ Searches Google for latest docs using Serper API  
✅ Supports:
- LangChain
- LlamaIndex
- OpenAI

✅ Returns plain text from documentation pages  
✅ Handles timeouts and search failures gracefully  

---

## 📁 Project Structure

mcp-docs-server/ ├── server.js # The MCP server implementation ├── .env # Environment variables └── README.md # This file

---

## ⚙️ Setup

1. **Clone the repository**  
   ```bash
   git clone <repo-url>
   cd mcp-docs-server

2. Install dependencies

   npm install
   

3. Create a .env file

   SERPER_API_KEY=your_serper_api_key
   

4. Run the MCP server

   node server.js
   

---

🛠️ Usage

The server exposes one tool:

get_docs

• Description: Searches the latest documentation for a given query and library.

• Arguments:

  • query (string): e.g. "Chroma DB"
  • library (string): one of "langchain", "llama-index", "openai"

• Returns: Plain text content fetched from the search results.

Example usage from a compatible MCP client:

node client.js path/to/server.js

---

📦 Dependencies

• @modelcontextprotocol/sdk
• dotenv
• httpx
• jsdom

---

📝 Notes

• The MCP server uses the Serper API for Google search.
• To get a Serper API key, visit serper.dev.

---

🤝 Contributing

Feel free to open issues or submit pull requests!

---

📄 License

MIT License.

---

🔗 Links

• Model Context Protocol (MCP)
• Serper API