MCP Conversation Server

What is MCP Conversation Server?

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

How to use MCP Conversation Server?

Install the package via npm (npm install mcp-conversation-server), then configure it using a YAML file (config/models.yaml) that specifies your OpenRouter API key, persistence path, model definitions, and default model. In your application, import ConversationServer from the package, instantiate it with the loaded config, and call run(). The server exposes MCP tools (create-conversation, send-message, list-conversations) and resources (conversation://{id}, conversation://list).

Key features of MCP Conversation Server

• Full MCP protocol compliance with resource management and tool-based interaction
• Integration with all OpenRouter models including Claude 3 Opus, Sonnet, and Llama 2 70B
• Real-time streaming responses with chunked handling and token counting
• Create and manage multiple conversations with message history tracking
• Token usage monitoring and model context window management
• File system persistence with configurable storage location and automatic state management
• Conversation filtering and search capabilities

Use cases of MCP Conversation Server

• Building a chatbot that can switch between different LLM providers via OpenRouter
• Creating a conversation-based tool or agent that persists discussions across sessions
• Developing an application that needs to track token usage and context windows per conversation
• Enabling streaming interactions with language models for real-time response display

FAQ from MCP Conversation Server

What does MCP Conversation Server replace?

This server replaces manual OpenRouter API calls by providing a standardized MCP interface. It manages conversations, token counting, and persistence so applications don’t need to implement those features themselves.

What are the runtime dependencies?

The server requires Node.js, an OpenRouter API key (configured in config/models.yaml), and the mcp-conversation-server npm package. No environment variables are needed since all configuration is done through the YAML file.

Where does conversation data live?

Conversation state is persisted to the file system. The default directory is ./conversations but can be changed via the persistence.path setting in the YAML configuration file.

Does the server support streaming?

Yes. The send-message tool accepts a stream boolean parameter to enable real-time streaming responses, and the server includes chunked response handling and token counting for streaming.

What authentication does MCP Conversation Server use?

Authentication is handled through the OpenRouter API key provided in the YAML configuration. The server does not expose its own authentication layer; it relies on OpenRouter’s key validation for model access.