MCP Chat Adapter

What is MCP Chat Adapter?

MCP Chat Adapter is an MCP (Model Context Protocol) server that provides a clean interface for LLMs to use chat completion capabilities through the MCP protocol. It acts as a bridge between an LLM client and any OpenAI‑compatible API, and is designed specifically for chat models (not text completions).

How to use MCP Chat Adapter?

Install via npm and configure environment variables (OPENAI_API_KEY, OPENAI_API_BASE, CONVERSATION_DIR) in your MCP client’s mcp.json file. The server exposes tools for creating conversations, chatting, listing, getting, and deleting conversations. Conversations are stored locally in the directory specified by CONVERSATION_DIR.

Key features of MCP Chat Adapter

• Built with FastMCP for a robust, clean implementation.
• Provides tools for conversation management and chat completion.
• Supports conversation persistence with local file storage.
• Compatible with OpenAI and any OpenAI‑compatible API.
• Configurable model, system prompt, and generation parameters.
• Proper error handling and timeouts.

Use cases of MCP Chat Adapter

• Allow an LLM like Claude to spin off and maintain multiple background conversations with other models.
• Continue an existing conversation across different contexts by specifying its conversation_id.
• Manually edit or review conversation history by modifying files in the CONVERSATION_DIR.
• Integrate with any tool that speaks MCP to provide chat completion capabilities.

FAQ from MCP Chat Adapter

What dependencies or runtime does it require?

Node.js (via npx) and an API key for an OpenAI‑compatible service. The server is published on npm and runs with one command.

Where are conversations stored?

Conversations are stored as local files in the directory set by the CONVERSATION_DIR environment variable (default ./convos). The server must be restarted to see manual edits.

Does it support text completions?

No. The server is focused on chat models only and does not support text completions.

Which APIs can I use?

Any API that is compatible with OpenAI’s chat completion endpoint. For example, OpenRouter is explicitly mentioned in the README’s configuration.

What are the storage limits?

The MAX_CONVERSATIONS environment variable (default 1000) sets the maximum number of conversations stored. Old conversations are not automatically evicted beyond this limit.