MCP Sse Client Python

🚀 MCP SSE Client

The MCP SSE Client is a Python library designed to interact with MCP services that support Server-Sent Events (SSE). It allows you to invoke tools in a simple way and integrates multiple LLM providers (such as OpenAI, Anthropic, and Ollama) for AI-driven tool selection.

🚀 Quick Start

1. Clone the repository:

   git clone https://github.com/zanetworker/mcp-sse-client-python.git
   cd mcp-sse-client-python
   pip install -e .
   

2. Run the Streamlit application:

   cd mcp-streamlit-app
   streamlit run app.py
   

✨ Features

1. Simple MCP Client

client = MCPClient(endpoint)

• endpoint: The URL of the MCP service (must be http or https).

Methods

async list_tools() -> List[ToolDef]

List all available tools on the MCP service.

async invoke_tool(tool_name: str, kwargs: Dict[str, Any]) -> ToolInvocationResult

Invoke a specific tool by name and pass parameters.

2. LLM-Driven Tool Selection

Supports multiple LLM providers for automatically selecting and executing tools based on user queries:

OpenAI Bridge

bridge = OpenAIBridge(mcp_client, api_key, model="gpt-4o")

Anthropic Bridge

bridge = AnthropicBridge(mcp_client, api_key, model="claude-3-opus-20240229")

Ollama Bridge

bridge = OllamaBridge(mcp_client, model="llama3", host=None)

Common Methods

async process_query(query: str) -> Dict[str, Any]

Process user queries, perform tool selection and execution through the LLM.

3. Command-Line Interface (CLI)

Provides a powerful CLI tool that supports the following operations:

• Connect to any MCP endpoint.
• Select an LLM among OpenAI, Anthropic, and Ollama.
• View available tools and their parameters.
• Test tools using natural language.
• Visualize tool selection reasoning and results.

4. Interactive Test UI

Includes a Streamlit application that provides a user-friendly interface:

• Connect to any MCP endpoint.
• Select an LLM among OpenAI, Anthropic, and Ollama.
• View available tools and their parameters.
• Test tools using natural language.
• Visualize tool selection reasoning and results.

📦 Installation

Install from Source

git clone https://github.com/zanetworker/mcp-sse-client-python.git
cd mcp-sse-client-python
pip install -e .

Install with pip (Once Published)

pip install mcp-sse-client

Supported LLM Providers

• OpenAI: Supports models such as GPT-4o, GPT-4, and GPT-3.5-Turbo.
• Anthropic: Supports models such as Claude 3 Opus, Claude 3 Sonnet, and Claude 3 Haiku.
• Ollama: Supports locally hosted models such as Llama 3 and Mistral.

📚 Documentation

MCPClient

client = MCPClient(endpoint)

• endpoint: The URL of the MCP service (must be http or https).

Methods

async list_tools() -> List[ToolDef]

List available tools.

async invoke_tool(tool_name: str, kwargs: Dict[str, Any]) -> ToolInvocationResult

Invoke a specific tool by name and pass parameters.

LLM Bridge

Provides bridge classes for multiple LLM providers for unified interface calls:

OpenAI Bridge

bridge = OpenAIBridge(mcp_client, api_key, model="gpt-4o")

Anthropic Bridge

bridge = AnthropicBridge(mcp_client, api_key, model="claude-3-opus-20240229")

Ollama Bridge

bridge = OllamaBridge(mcp_client, model="llama3", host=None)

Requirements

• Python version: 3.8+
• Additional dependencies:
  • requests
  • sseclient-py
  • langchain
  • openai
  • anthropic
  • llama-cp

Project Structure

mcp-sse-client-python/
├── mcp_sse_client/
│   ├── client.py
│   ├── tools.py
│   └── bridges/
│       ├── openai_bridge.py
│       ├── anthropic_bridge.py
│       └── olalla_bridge.py
└── mcp-streamlit-app/
    ├── app.py
    └── templates/
        └── index.html

💻 Usage Examples

Basic Usage

from mcp_sse_client.client import MCPClient
from mcp_sse_client.bridges.openai_bridge import OpenAIBridge

# Initialize the MCP client
client = MCPClient("http://localhost:8000")

# Get available tools
tools = client.list_tools()
print(tools)

# Invoke a tool by name
result = await client.invoke_tool("tool_name", {"param": "value"})
print(result)

LLM-Driven Tool Selection

from mcp_sse_client.bridges.openai_bridge import OpenAIBridge

# Initialize the OpenAI bridge
bridge = OpenAIBridge(mcp_client, api_key="your-api-key", model="gpt-4")

# Process user queries
result = await bridge.process_query("Your query content")
print(result)

Project Goals

The MCP SSE Client aims to simplify interactions with MCP services while providing powerful LLM integration capabilities. Through a unified interface and flexible tool invocation mechanism, it helps developers quickly build intelligent applications.

Contribution Guidelines

Contributions are welcome! Please refer to CONTRIBUTING.md to learn how to participate in the development.