Model Context Protocol (MCP) Server

A server implementation of the Model Context Protocol for managing context between AI models and applications.

What is MCP?

The Model Context Protocol (MCP) is a protocol for managing context between AI models and applications. It provides a standardized way to:

1. Create and manage sessions for user interactions
2. Store and retrieve context data
3. Process context with AI models
4. Merge and summarize contexts
5. Query models with context-aware prompts

Features

• Session management
• Context storage and retrieval
• Context processing with AI models
• Context merging and summarization
• Model querying with context

Getting Started

1. Clone this repository
2. Install dependencies:

   npm install
   

3. Configure environment variables in .env:

   PORT=3000
   API_KEY=your_api_key_here
   MODEL_ENDPOINT=http://localhost:8000/v1/completions
   CONTEXT_DB_PATH=./data/context_db.json
   MODEL_TYPE=mistral
   

4. Start the server:

   npm start
   

API Endpoints

Sessions

• POST /api/session - Create a new session
• GET /api/session/:sessionId - Get a session by ID
• PATCH /api/session/:sessionId - Update a session
• DELETE /api/session/:sessionId - Delete a session
• GET /api/session/:sessionId/contexts - Get all contexts for a session

Contexts

• POST /api/context - Create a new context
• GET /api/context/:contextId - Get a context by ID
• PATCH /api/context/:contextId - Update a context
• DELETE /api/context/:contextId - Delete a context
• POST /api/context/merge - Merge multiple contexts
• POST /api/context/:contextId/summarize - Summarize a context

Models

• POST /api/model/query - Query the model with a prompt and context
• POST /api/model/process - Process context with a model

Authentication

All API endpoints are protected with API key authentication. Include your API key in the Authorization header:

Authorization: Bearer your_api_key_here

Data Storage

By default, context data is stored in a JSON file at ./data/context_db.json. You can configure a different path in the .env file.

Model Integration

The server is designed to work with any AI model that supports a compatible API. Configure the model endpoint in the .env file.

Supported Model Types

The server supports different model types through the MODEL_TYPE environment variable:

• mistral - For Mistral AI models
• openai - For OpenAI models
• anthropic - For Anthropic models
• generic - For other model providers

SSL Certificate Verification

For development purposes, SSL certificate verification is disabled by default. In production environments, you should:

1. Either provide proper certificates
2. Or remove the rejectUnauthorized: false option in modelManager.js

Example Usage

Create a Session

curl -X POST http://localhost:3000/api/session \
  -H "Authorization: Bearer your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{"metadata": {"user": "example_user"}}'

Create a Context

curl -X POST http://localhost:3000/api/context \
  -H "Authorization: Bearer your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "sessionId": "session_123",
    "data": {
      "conversation": [
        {"role": "user", "content": "Hello, how are you?"},
        {"role": "assistant", "content": "I'm doing well, thank you for asking!"}
      ]
    }
  }'

Query a Model with Context

curl -X POST http://localhost:3000/api/model/query \
  -H "Authorization: Bearer your_api_key_here" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "What was my last question?",
    "contextId": "ctx_123",
    "sessionId": "session_123"
  }'

License

MIT