MCP Conversation Server

A Model Context Protocol (MCP) server implementation for managing conversations with OpenRouter's language models. This server provides a standardized interface for applications to interact with various language models through a unified conversation management system.

Features

• MCP Protocol Support
  • Full MCP protocol compliance
  • Resource management and discovery
  • Tool-based interaction model
  • Streaming response support
  • Error handling and recovery

• OpenRouter Integration
  • Support for all OpenRouter models
  • Real-time streaming responses
  • Automatic token counting
  • Model context window management
  • Available models include:
    • Claude 3 Opus
    • Claude 3 Sonnet
    • Llama 2 70B
    • And many more from OpenRouter's catalog

• Conversation Management
  • Create and manage multiple conversations
  • Support for system messages
  • Message history tracking
  • Token usage monitoring
  • Conversation filtering and search

• Streaming Support
  • Real-time message streaming
  • Chunked response handling
  • Token counting

• File System Persistence
  • Conversation state persistence
  • Configurable storage location
  • Automatic state management

Installation

npm install mcp-conversation-server

Configuration

Configuration

All configuration for the MCP Conversation Server is now provided via YAML. Please update the `config/models.yaml` file with your settings. For example:

# MCP Server Configuration
openRouter:
  apiKey: "YOUR_OPENROUTER_API_KEY"  # Replace with your actual OpenRouter API key.

persistence:
  path: "./conversations"  # Directory for storing conversation data.

models:
  # Define your models here
  'provider/model-name':
    id: 'provider/model-name'
    contextWindow: 123456
    streaming: true
    temperature: 0.7
    description: 'Model description'

# Default model to use if none specified
defaultModel: 'provider/model-name'

Server Configuration

The MCP Conversation Server now loads all its configuration from the YAML file. In your application, you can load the configuration as follows:

const config = await loadModelsConfig(); // Loads openRouter, persistence, models, and defaultModel settings from 'config/models.yaml'

*Note: Environment variables are no longer required as all configuration is provided via the YAML file.*

Usage

Basic Server Setup

import { ConversationServer } from 'mcp-conversation-server';

const server = new ConversationServer(config);
server.run().catch(console.error);

Available Tools

The server exposes several MCP tools:

1. create-conversation

{
       provider: 'openrouter',    // Provider is always 'openrouter'
       model: string,             // OpenRouter model ID (e.g., 'anthropic/claude-3-opus-20240229')
       title?: string;            // Optional conversation title
   }

2. send-message

{
       conversationId: string;  // Conversation ID
       content: string;         // Message content
       stream?: boolean;        // Enable streaming responses
   }

3. list-conversations

{
       filter?: {
           model?: string;      // Filter by model
           startDate?: string;  // Filter by start date
           endDate?: string;    // Filter by end date
       }
   }

Resources

The server provides access to several resources:

1. conversation://{id}

2. conversation://list

Development

Building

npm run build

Running Tests

npm test

Debugging

The server provides several debugging features:

1. Error Logging

2. MCP Inspector

npm run inspector

Use the MCP Inspector to:

3. Provider Validation

await server.providerManager.validateProviders();

Validates:

Troubleshooting

Common issues and solutions:

1. OpenRouter Connection Issues

2. Message Streaming Errors

3. File System Errors

Contributing

1. Fork the repository

2. Create a feature branch

3. Commit your changes

4. Push to the branch

5. Create a Pull Request

License

ISC License