Model-Context-Protocol

MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.

---

AI Agent System with MCP Gateway

---

Prerequisite

• Python 3.10 or higher installed.

• You must use the Python MCP SDK 1.2.0 or higher.

• Install uv package manager (Optional)

---

Quickstart

Clone the repository

```bash
$ git clone <clone-url>
```

Run the MCP servers

$ cd mcp-servers/<perticular-server-folder>

$ python3 -m venv .venv

$ source .venv/bin/activate # For Windows: .venv\Scripts\activate

$ pip install -r requirements.txt # or uv sync

$ python <python-file>.py

Run MCP Client to test the MCP server (Optional)

1. Setup

$ cd mcp-clients/simple-cient

$ python3 -m venv .venv

$ source .venv/bin/activate # For Windows: .venv\Scripts\activate

$ pip install -r requirements.txt # or uv sync

2. Replace the MCP Server endpoint as url value in the client.py.

3. Run the Client (Currently, only listing the tools.)

$ python client.py

Run the MCP gateway

$ cd mcp-gateway

$ python3 -m venv .venv

$ source .venv/bin/activate # For Windows: .venv\Scripts\activate

$ pip install -r requirements.txt

$ python gateway.py

Run the MCP Agent

1. Go to the MCP agent directory.

$ cd mcp-agent

2. Create a .env file

LLM_PROVIDER=gemini
GOOGLE_API_KEY=<gemini-api-key>

3. Create virtual environment and install dependencies.

$ python3 -m venv .venv

$ source .venv/bin/activate # For Windows: .venv\Scripts\activate

$ pip install -r requirements.txt

4. Run the agent.

$ python main.py

Ask queries in the agent and find the result

Play around with the servers and agents.

---

Add New MCP Servers

If you want to add new MCP Servers to the gateway. Add it's endpoint to SERVER_REGISTRY dictonary in gateway.py.

---

MCP Servers

MCP servers the repository contains are created using FastMCP package from MCP Python SDK.

• It uses Streamable HTTP transport.
• Its stateless.

MCP Gateway

An MCP Gateway is a crucial piece of infrastructure that sits between MCP clients (agents) and a multitude of MCP servers (tools). It acts as a centralized control plane, offering:

1. Unified Access: A single entry point to a wide range of tools.

2. Security: Centralized authentication, authorization, and secrets management.

3. Observability: Monitoring, logging, and tracing of all tool interactions.

4. Management: A central place to register, discover, and manage available tools.

Agent

Supported LLM provider