MCP Chat Adapter
About
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).
Details
- License
- MIT license
Explore
- Built with FastMCP for robust and clean implementation
- Provides tools for conversation management and chat completion
- Proper error handling and timeouts
- Supports conversation persistence with local storage
- Easy setup with minimal configuration
- Configurable model parameters and defaults
- Compatible with OpenAI and OpenAI-compatible APIs
These environment variables must be set for the server to function:
OPENAI_API_KEY=your-api-key # Your API key for OpenAI or compatible service
OPENAI_API_BASE=https://openrouter.ai/api/v1 # The base URL for the API (can be changed for compatible services)
You should also set the CONVERSATION_DIR environment variable to the directory where you want to store the conversation data. Use an absolute path.
The following environment variables are optional and have default values:
DEFAULT_MODEL=google/gemini-2.0-flash-001 # Default model to use if not specified
DEFAULT_SYSTEM_PROMPT="You are an unhelpful assistant." # Default system prompt
DEFAULT_MAX_TOKENS=50000 # Default maximum tokens for completion
DEFAULT_TEMPERATURE=0.7 # Default temperature setting
DEFAULT_TOP_P=1.0 # Default top_p setting
DEFAULT_FREQUENCY_PENALTY=0.0 # Default frequency penalty
DEFAULT_PRESENCE_PENALTY=0.0 # Default presence penalty
CONVERSATION_DIR=./convos # Directory to store conversation data
MAX_CONVERSATIONS=1000 # Maximum number of conversations to store
yarn install
For FastMCP cli run:
yarn cli
For FastMCP inspect run:
yarn inspect
```
An MCP (Model Context Protocol) server that provides a clean interface for LLMs to use chat completion capabilities through the MCP protocol. This server acts as a bridge between an LLM client and any OpenAI-compatible API. The primary use case is for chat models, as the server does not provide support for text completions.
Overview
The OpenAI Chat MCP Server implements the Model Context Protocol (MCP), allowing language models to interact with OpenAI's chat completion API in a standardized way. It enables seamless conversations between users and language models while handling the complexities of API interactions, conversation management, and state persistence.
Features
- Built with FastMCP for robust and clean implementation
- Provides tools for conversation management and chat completion
- Proper error handling and timeouts
- Supports conversation persistence with local storage
- Easy setup with minimal configuration
- Configurable model parameters and defaults
- Compatible with OpenAI and OpenAI-compatible APIs
Typical Workflow
The idea is that you can have Claude spin off and maintain multiple conversations with other models in the background. All conversations are stored in the CONVERSATION_DIR directory, which you should set in the env section of your mcp.json file.
It is possible to tell Claude either to create a new conversation, or to continue an existing one (identified by the integer conversation_id). You can continue with the old conversation even if you are starting fresh in a new context, although in that case you may want to tell Claude to read the old conversation before continuing using the get_conversation tool.
Note that you can also edit the conversations in the CONVERSATION_DIR directory manually. In this case, you may need to restart the server to see the changes.
Configuration
Required Environment Variables
These environment variables must be set for the server to function:
OPENAI_API_KEY=your-api-key # Your API key for OpenAI or compatible service
OPENAI_API_BASE=https://openrouter.ai/api/v1 # The base URL for the API (can be changed for compatible services)
You should also set the CONVERSATION_DIR environment variable to the directory where you want to store the conversation data. Use an absolute path.
Optional Environment Variables
The following environment variables are optional and have default values:
```sh
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
