Cross-LLM MCP Server

A Model Context Protocol (MCP) server that provides access to multiple Large Language Model (LLM) APIs including ChatGPT, Claude, and DeepSeek. This allows you to call different LLMs from within any MCP-compatible client and combine their responses.

Features

This MCP server offers five specialized tools for interacting with different LLM providers:

🤖 Individual LLM Tools

call-chatgpt

Call OpenAI's ChatGPT API with a prompt.

Input:

• prompt (string): The prompt to send to ChatGPT

• model (optional, string): ChatGPT model to use (default: gpt-4)

• temperature (optional, number): Temperature for response randomness (0-2, default: 0.7)

• max_tokens (optional, number): Maximum tokens in response (default: 1000)

Output:

• ChatGPT response with model information and token usage statistics

Example:

call-claude

Call Anthropic's Claude API with a prompt.

Input:

• prompt (string): The prompt to send to Claude

• model (optional, string): Claude model to use (default: claude-3-sonnet-20240229)

• temperature (optional, number): Temperature for response randomness (0-1, default: 0.7)

• max_tokens (optional, number): Maximum tokens in response (default: 1000)

Output:

• Claude response with model information and token usage statistics

call-deepseek

Call DeepSeek API with a prompt.

Input:

• prompt (string): The prompt to send to DeepSeek

• model (optional, string): DeepSeek model to use (default: deepseek-chat)

• temperature (optional, number): Temperature for response randomness (0-2, default: 0.7)

• max_tokens (optional, number): Maximum tokens in response (default: 1000)

Output:

• DeepSeek response with model information and token usage statistics

🔄 Combined Tools

call-all-llms

Call all available LLM APIs (ChatGPT, Claude, DeepSeek) with the same prompt and get combined responses.

Input:

• prompt (string): The prompt to send to all LLMs

• temperature (optional, number): Temperature for response randomness (0-2, default: 0.7)

• max_tokens (optional, number): Maximum tokens in response (default: 1000)

Output:

• Combined responses from all LLMs with individual model information and usage statistics

• Summary of successful responses and total tokens used

Example:

call-llm

Call a specific LLM provider by name.

Input:

• provider (string): The LLM provider to call ("chatgpt", "claude", or "deepseek")

• prompt (string): The prompt to send to the LLM

• model (optional, string): Model to use (uses provider default if not specified)

• temperature (optional, number): Temperature for response randomness (0-2, default: 0.7)

• max_tokens (optional, number): Maximum tokens in response (default: 1000)

Output:

• Response from the specified LLM with model information and usage statistics

Installation

1. Clone this repository:

2. Install dependencies:

3. Build the project:

Getting API Keys

OpenAI/ChatGPT

1. Visit OpenAI Platform

2. Sign up or log in to your account

3. Create a new API key

4. Add it to your .env file as OPENAI_API_KEY

Anthropic/Claude

1. Visit Anthropic Console

2. Sign up or log in to your account

3. Create a new API key

4. Add it to your .env file as ANTHROPIC_API_KEY

DeepSeek

1. Visit DeepSeek Platform

2. Sign up or log in to your account

3. Create a new API key

4. Add it to your .env file as DEEPSEEK_API_KEY

Usage

Configuring Claude Desktop

Add the following configuration to your Claude Desktop MCP settings:

Replace the paths and API keys with your actual values:

• Update the args path to point to your build/index.js file

• Update the cwd path to your project directory

• Add your actual API keys to the env section

Running the Server

The server runs automatically when configured in Claude Desktop. You can also run it manually:

The server runs on stdio and can be connected to any MCP-compatible client.

Example Queries

Here are some example queries you can make with this MCP server:

Call ChatGPT

Call Claude

Call All LLMs

Call Specific LLM

Use Cases

1. Multi-Perspective Analysis

Use call-all-llms to get different perspectives on the same topic from multiple AI models.

2. Model Comparison

Compare responses from different LLMs to understand their strengths and weaknesses.

3. Redundancy and Reliability

If one LLM is unavailable, you can still get responses from other providers.

4. Cost Optimization

Choose the most cost-effective LLM for your specific use case.

5. Quality Assurance

Cross-reference responses from multiple models to validate information.

Configuration

Claude Desktop Setup

The recommended way to use this MCP server is through Claude Desktop with environment variables configured directly in the MCP settings:

Environment Variables

The server reads the following environment variables:

• OPENAI_API_KEY: Your OpenAI API key

• ANTHROPIC_API_KEY: Your Anthropic API key

• DEEPSEEK_API_KEY: Your DeepSeek API key

• DEFAULT_CHATGPT_MODEL: Default ChatGPT model (default: gpt-4)

• DEFAULT_CLAUDE_MODEL: Default Claude model (default: claude-3-sonnet-20240229)

• DEFAULT_DEEPSEEK_MODEL: Default DeepSeek model (default: deepseek-chat)

API Endpoints

This MCP server uses the following API endpoints:

• OpenAI: https://api.openai.com/v1/chat/completions

• Anthropic: https://api.anthropic.com/v1/messages

• DeepSeek: https://api.deepseek.com/v1/chat/completions

Error Handling

The server includes comprehensive error handling with detailed messages:

Missing API Key

Invalid API Key

Rate Limiting

Payment Issues

Network Issues

Supported Models

ChatGPT Models

• gpt-4

• gpt-4-turbo

• gpt-3.5-turbo

• And other OpenAI models

Claude Models

• claude-3-sonnet-20240229

• claude-3-opus-20240229

• claude-3-haiku-20240307

• And other Anthropic models

DeepSeek Models

• deepseek-chat

• deepseek-coder

• And other DeepSeek models

Project Structure

Dependencies

• @modelcontextprotocol/sdk - MCP SDK for server implementation

• superagent - HTTP client for API requests

• zod - Schema validation for tool parameters

Development

Building the Project

Adding New LLM Providers

To add a new LLM provider:

1. Add the provider type to src/types.ts

2. Implement the client in src/llm-clients.ts

3. Add the tool to src/index.ts

4. Update the callAllLLMs method to include the new provider

Troubleshooting

Common Issues

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

Debug Mode

For debugging, you can run the server directly:

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Support

If you encounter any issues or have questions, please:

1. Check the troubleshooting section above

2. Review the error messages for specific guidance

3. Ensure your API keys are properly configured

4. Verify your network connectivity