MCP

sample implementation of client and server using the MCP Python SDK and Google's Gemini LLM

Features

• Client Components

  • FastAPI backend for handling requests
  • Streamlit-based web user interface
  • Integration with Google's Gemini LLM
  • Configurable MCP server connections

• Server Components

  • Document search using vector embeddings
  • Persistent vector storage using Chroma DB
  • MCP servers & tools

Prerequisites

• Python 3.12 or higher
• Environment variables set up (see .env.example files)
• Required API keys:
  • Google Gemini API key
  • Serper API key (for web search)

Configuration

The client can be configured using mcp_servers_config.json:

{
    "mcpServers": {
        "server_name": {
            "command": "command_to_run_server",
            "args": ["arg1", "arg2"]
        }
    }
}