MCP-SERVER-CLIENT-SSE

147MahirJawad/MCP-SERVER-CLIENT-SSE

3.2

If you are the rightful owner of MCP-SERVER-CLIENT-SSE and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

This project implements a client-server system using Server-Sent Events (SSE) for real-time communication between a Gemini AI-powered client and a tool-providing server.

Tools

  1. Web Search

    Performs web searches using Tavily API.

  2. Run Command

    Executes shell commands in a workspace directory.

  3. Add Numbers

    Performs basic arithmetic operations.

MCP SSE Client-Server System

Overview

This project implements a client-server system using Server-Sent Events (SSE) for real-time communication between a Gemini AI-powered client and a tool-providing server. The system allows users to interact with various tools (like web search and command execution) through natural language queries processed by Gemini AI.

Features

  • SSE-based Communication: Real-time, bidirectional communication between client and server
  • Gemini AI Integration: Natural language processing for tool selection and query understanding
  • Tool Ecosystem: Includes tools for:
    • Web search (via Tavily API)
    • Shell command execution
    • Basic arithmetic operations
  • Asynchronous Architecture: Built with Python's asyncio for efficient I/O operations

Prerequisites

  • Python 3.7+
  • Required environment variables:
    • GEMINI_API_KEY - For Gemini AI access
    • TAVILY_API_KEY - For web search functionality

Installation

  1. Clone the repository:

    git clone https://github.com/147MahirJawad/mcp-server-client-sse.git
cd mcp-sse-system

  2. Install dependencies:

    pip install -r requirements.txt

  3. Create a .env file with your API keys:

    GEMINI_API_KEY=your_gemini_key
TAVILY_API_KEY=your_tavily_key

Usage

Running the Server

Start the MCP server with:

python server_sse.py --host 0.0.0.0 --port 8081

Optional arguments:

  • --host: Server host (default: 0.0.0.0)
  • --port: Server port (default: 8081)

Running the Client

Connect to the server with:

python client_sse.py http://localhost:8081/sse

Once connected, you can:

  • Enter natural language queries
  • The system will automatically select and use appropriate tools
  • Type 'exit' to quit

Available Tools

  1. Web Search: Performs web searches using Tavily API

    • Example: "Search for latest AI news"

  2. Run Command: Executes shell commands in a workspace directory

    • Example: "List files in the workspace"

  3. Add Numbers: Performs basic arithmetic

    • Example: "What's 123 plus 456?"

Architecture

Client (Gemini AI) ↔ SSE Transport ↔ MCP Server ↔ Tools
  • Client: Uses Gemini AI to interpret user queries and select tools
  • SSE Transport: Handles real-time communication
  • MCP Server: Manages tool execution and responses
  • Tools: Various functionality endpoints

Configuration

  • Workspace directory can be modified in server_sse.py (DEFAULT_WORKSPACE)
  • Additional tools can be added by decorating functions with @mcp.tool()

Troubleshooting

  • Ensure all API keys are set in .env
  • Verify server is running before starting client
  • Check port availability if connection fails

License