Simple MCP Client

🚀 Simple MCP Client Usage Documentation

The Simple MCP Client is a tool designed to interact with various models and services, supporting both local and cloud-based machine learning models. Here are the core features and configuration instructions for this client.

🚀 Quick Start

To start the client, run the following command in the terminal:

python -m simple_mcp_client.cli

You can also specify the path to the configuration file:

python -m simple_mcp_client.cli --config path/to/config.json

📚 Documentation

Configuration File Format

The configuration file is a JSON file containing the following fields:

{
  "log_level": "INFO", // Optional values: DEBUG, INFO, WARNING, ERROR, CRITICAL
  "llm": {
    "provider": "ollama", // Supported LLM providers: ollama, openai, deepseek, openrouter, litellm
    "model": "gpt-4-turbo", // Model name, depending on the provider
    "api_key": "your_api_key", // Required API key
    // Refer to the corresponding provider's documentation for other parameters
  },
  "mcp_servers": [
    {
      "url": "http://localhost:8000",
      "name": "local_mcp"
    },
    {
      "url": "https://api.example.com",
      "name": "remote_mcp"
    }
  ],
  "default_mcp_server": "local_mcp" // The default MCP server to use
}

Command-Line Options

Here are some commonly used command-line options:

• --help: Display help information.
• --version: Display the client version.
• --log-level <LEVEL>: Set the log level.
• --config <PATH>: Specify the path to the configuration file.

Environment Variables

You can configure the following through environment variables:

• MCP_LOG_LEVEL: Set the log level (default: INFO).
• OPENAI_API_KEY: OpenAI API key.
• DEEPSEEK_API_KEY: DeepSeek API key.
• OPENROUTER_API_KEY: OpenRouter API key.
• ANTHROPIC_API_KEY: Anthropic API key.
• GOOGLE_API_KEY: Google API key.
• MISTRAL_API_KEY: Mistral API key.
• COHERE_API_KEY: Cohere API key.
• AZURE_API_KEY: Azure API key.

LLM Provider Support

The client supports the following LLM providers:

1. Ollama: Run open-source models locally.
2. OpenAI: An API-based LLM provider for accessing OpenAI's models.
3. DeepSeek: An API-based LLM provider offering multiple models.
4. OpenRouter: A unified API interface supporting multiple LLM providers, providing a wide range of model choices.
5. LiteLLM: A unified interface supporting multiple providers such as OpenAI, Anthropic, Google, and Azure.

DeepSeek Configuration

• An API key is required, which can be provided through the configuration file or the environment variable DEEPSEEK_API_KEY.
• The default API URL is https://api.deepseek.com/v1.

OpenRouter Configuration

• An API key is required, which can be provided through the configuration file or the environment variable OPENROUTER_API_KEY.
• The default API URL is https://openrouter.ai/api/v1.
• Additional parameters are supported:
  • http_referer: The default value is "https://github.com/simple_mcp_client".
  • app_name: The default value is "Simple MCP Client".

LiteLLM Configuration

It supports multiple model providers. Here is an example configuration:

{
  "llm": {
    "provider": "litellm",
    "model": "anthropic/claude-3-opus-20240229", // Example model name
    "api_key": "YOUR_ANTHROPIC_API_KEY", // Required API key
    "other_params": {
      "temperature": 0.7,
      "max_tokens": 4096,
      "request_timeout": 60.0
    }
  }
}

For OpenAI models:

{
  "llm": {
    "provider": "litellm",
    // Other relevant configuration content should be here according to the original text
  }
}