Model Context Protocol (MCP) Client

Unleash the full potential of Large Language Models (LLMs) by connecting them to a universe of external tools! This project presents a robust, multi-server client built on the Model Context Protocol (MCP), enabling your LLM to discover, understand, and invoke custom functionalities. Seamlessly integrate tools via STDIO or SSE, orchestrate complex tasks with Google's powerful Gemini models, and chat across multi-turn conversations.

---

🚀 Core Concept

Modern LLMs are incredibly powerful but are often limited to the information they were trained on. The Model Context Protocol (MCP) is a specification that allows an LLM to discover and invoke external tools securely and efficiently.

This project implements an MCP client that acts as an intelligent orchestrator. It takes a user's query, consults an LLM (like Gemini) to form a plan, and then calls the necessary tools hosted on various MCP servers to gather information and execute tasks — all while supporting context-aware, multi-turn interactions.

---

🏛️ Architecture

The architecture is designed for scalability and flexibility, allowing you to add new tools and servers without modifying the core client logic.

---

✨ Features

• ✅ Multi-Turn Conversation Support: Maintain memory across chat turns for smarter, context-aware interactions.
• 🌐 Multi-Server Connectivity: Connect to multiple MCP servers simultaneously.
• 🔌 Flexible Transports: Supports both stdio for local process-based servers and sse (Server-Sent Events) for remote HTTP-based servers.
• 🧠 LLM-Powered Orchestration: Uses Gemini to interpret complex requests, plan tool usage, and deliver intelligent responses.
• 🖥️ Optional Web UI (FastAPI): RESTful API with Swagger/ReDoc docs for intuitive browsing and usage.
• 🔍 Tool Discovery: Dynamically detects and loads available tools from all connected servers.
• ⚙️ UI-Based Config & Logs: Configure API keys, tools, and servers from the web interface. Easily view logs and debug sessions.
• ⚡ Async Architecture: Built on asyncio for non-blocking performance.
• 🧾 Readable Configuration: JSON-based server configs and .env for secure credential management.
• 🛠 Dev-Friendly Logs: Inspect agent reasoning and tool calls for full transparency.

---

🏁 Getting Started

1. Prerequisites

• Python 3.10+
• Google AI Studio API key

2. Installation

git clone https://github.com/Abdulabin/Model-Context-Protocol.git
cd Model-Context-Protocol
pip install -r requirements.txt

3. Configuration

.env File (API Key):

GEMINI_API_KEY="YOUR_API_KEY_HERE"

Tool Server Config (mcp-servers-configs.json):

{
  "mcpServers": {
    "arxiv_server": {
      "disabled": false,
      "transport_type": "stdio",
      "command": "python",
      "args": ["mcp_servers/arxiv_server.py"]
    },
    "remote_tool_server": {
      "disabled": true,
      "transport_type": "sse",
      "url": "http://localhost:8001/mcp"
    }
  }
}

---

🚀 Usage

CLI Usage 💻

python mcp_client.py

Then ask questions like:

Find 3 recent papers on "Large Language Models" and summarize the first one.

The client understands the multi-step process, delegates tasks to appropriate tools, and returns an answer — all with memory across interactions.

---

Web Interface 🌐

Launch the UI with:

uvicorn app:app

Open in Browser:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Web UI Capabilities:

• 🧠 Chat with memory (multi-turn conversation)
• 🛠 Add/Edit tool servers
• 🔐 Set API keys and model configs
• 📜 View detailed logs of interactions

---

🧾 Tool Config JSON Schema (UI Compatible)

Supports both stdio and sse. Example UI-loadable config:

[
  {
    "mcpServers": {
      "LocalScript": {
        "disabled": false,
        "command": "python3",
        "args": ["script.py"],
        "transport_type": "stdio"
      },
      "WeatherAPI": {
        "disabled": true,
        "transport_type": "sse",
        "url": "http://api.weather.com/sse",
        "timeout": 30
      }
    }
  }
]

---

🧭 Pages in the Web App

Page	Purpose
🏠 Home	Introduction and overview
⚙️ Settings	Add API keys, select models, manage tools
💬 Chat	Chat with the LLM + tools (multi-turn)
📜 Logs	Debug and monitor tool calls and plans

---

🎯 Ideal Use-Cases

• 🤖 AI Assistants with real-time tools
• 📚 Academic research bots using APIs
• 🧪 AI development environments
• 🎓 Educational demos combining LLMs + tooling

---

✅ Summary

MCP-Client bridges LLMs and real-world tools with seamless configuration, extensible architecture, and support for intelligent, multi-turn conversations. Whether you're building smart agents, research assistants, or developer tools — it's your gateway to next-level AI integration.

Connect your tools. Configure your model. Start chatting. 🚀