Skip to content

example-usage.md

Latest commit

 

History

History
223 lines (158 loc) · 4.8 KB

example-usage.md

File metadata and controls

223 lines (158 loc) · 4.8 KB

Cross-LLM MCP Server Usage Examples

This document provides examples of how to use the Cross-LLM MCP Server with different MCP clients.

Configuration

1. Set up your environment file

Copy the example environment file and add your API keys:

cp env.example .env

Edit .env with your actual API keys:

# OpenAI/ChatGPT API Key
OPENAI_API_KEY=sk-your-openai-key-here

# Anthropic/Claude API Key
ANTHROPIC_API_KEY=sk-ant-your-anthropic-key-here

# DeepSeek API Key
DEEPSEEK_API_KEY=sk-your-deepseek-key-here

# Default models (optional)
DEFAULT_CHATGPT_MODEL=gpt-4
DEFAULT_CLAUDE_MODEL=claude-3-sonnet-20240229
DEFAULT_DEEPSEEK_MODEL=deepseek-chat

# Hugging Face Inference Router (Hub model ids; token needs Inference access)
HF_TOKEN=hf_your_token_here
# DEFAULT_HUGGINGFACE_MODEL=Qwen/Qwen2.5-7B-Instruct

2. Build the server

npm run build

Usage with Different MCP Clients

Claude Desktop

  1. Open Claude Desktop
  2. Go to Settings → Model Context Protocol
  3. Add a new server with these settings:
    • Name: Cross-LLM
    • Command: node
    • Arguments: [path-to-your-project]/build/index.js
    • Working Directory: [path-to-your-project]

Example Conversations

Single LLM Call

User: "Call ChatGPT to explain quantum computing"

Claude: I'll use the Cross-LLM MCP server to call ChatGPT for you.

{
  "tool": "call-chatgpt",
  "arguments": {
    "prompt": "Explain quantum computing in simple terms",
    "temperature": 0.7,
    "max_tokens": 500
  }
}

Hugging Face (Inference Router)

{
  "tool": "call-huggingface",
  "arguments": {
    "prompt": "What is 2+2? Reply with one word.",
    "model": "Qwen/Qwen2.5-7B-Instruct",
    "max_tokens": 50
  }
}

Multi-LLM Comparison

User: "Get responses from all LLMs about the future of AI"

Claude: I'll call all available LLMs to get different perspectives.

{
  "tool": "call-all-llms",
  "arguments": {
    "prompt": "What do you think the future of artificial intelligence will look like in the next 10 years?",
    "temperature": 0.8,
    "max_tokens": 800
  }
}

Specific Model Selection

User: "Ask Claude to write a poem about technology"

Claude: I'll call the Claude API specifically for this request.

{
  "tool": "call-claude",
  "arguments": {
    "prompt": "Write a short poem about the impact of technology on modern life",
    "model": "claude-3-sonnet-20240229",
    "temperature": 0.9,
    "max_tokens": 300
  }
}

Advanced Usage

Custom Model Selection

You can specify different models for each provider:

{
  "tool": "call-llm",
  "arguments": {
    "provider": "chatgpt",
    "prompt": "Explain machine learning algorithms",
    "model": "gpt-4-turbo",
    "temperature": 0.5,
    "max_tokens": 1000
  }
}

Temperature Control

Adjust creativity vs consistency:

  • Low temperature (0.1-0.3): More focused, consistent responses
  • Medium temperature (0.4-0.7): Balanced creativity and consistency
  • High temperature (0.8-1.0): More creative, varied responses

Token Management

Control response length:

  • Short responses: 100-300 tokens
  • Medium responses: 500-800 tokens
  • Long responses: 1000+ tokens

Error Handling

The server provides clear error messages for common issues:

Missing API Key

**ChatGPT Error:** OpenAI API key not configured

Network Issues

**Claude Error:** Claude API error: Network timeout

Rate Limiting

**DeepSeek Error:** DeepSeek API error: Rate limit exceeded

Best Practices

  1. Start with individual calls to test each LLM before using call-all-llms
  2. Use appropriate temperature for your use case
  3. Monitor token usage to manage costs
  4. Handle errors gracefully - one LLM failure shouldn't stop your workflow
  5. Compare responses to understand different model strengths

Troubleshooting

Server won't start

  • Check that all dependencies are installed: npm install
  • Verify the build was successful: npm run build
  • Ensure the .env file exists and has valid API keys

API errors

  • Verify your API keys are correct and active
  • Check your API usage limits and billing status
  • Ensure you're using supported model names

No responses

  • Check that at least one API key is configured
  • Verify network connectivity
  • Look for error messages in the response

Integration Examples

With Claude Desktop

The server integrates seamlessly with Claude Desktop, allowing you to:

  • Call other LLMs while chatting with Claude
  • Compare responses from different models
  • Use specialized models for specific tasks

With Other MCP Clients

Any MCP-compatible client can use this server to:

  • Access multiple LLM providers
  • Get diverse perspectives on topics
  • Build more robust AI applications
⚡