MCP-Client

An async MCP Client built with Gemini and Python, supporting Streamable HTTP and STDIO transports for tool discovery and execution through an MCP server.

MCP Client

A Python-based MCP Client that connects a Gemini model to an MCP Server.
It discovers tools exposed by the server, converts them into Gemini-compatible function declarations, and enables interactive terminal chat where Gemini can invoke server tools automatically.

This client supports two transport modes:

• Streamable HTTP
• STDIO

---

Features

• Connects to an MCP server over HTTP or STDIO
• Loads configuration from a .env file
• Uses Gemini as the LLM backend
• Dynamically reads MCP tools from the server
• Maps MCP tool schemas into Gemini function declarations
• Lets Gemini call tools during chat
• Runs fully asynchronously with asyncio

---

How It Works

1. The client connects to an MCP server.
2. It fetches the list of available tools using list_tools().
3. It converts those tools into Gemini function declarations.
4. The user types a prompt in the terminal.
5. Gemini decides whether to answer directly or call a tool.
6. If a tool call is requested:
   • the client executes it through the MCP session
   • sends the result back to Gemini
   • prints the final assistant response

---

Requirements

Install dependencies:

pip install google-genai python-dotenv mcp