OpenAI MCP Server

Example Use Cases

• 1

  Call OpenAI models from an MCP host without directly handling HTTP API calls or keys.

• 2

  A lightweight local adapter to route model requests to a specific OpenAI model (configurable via env var).

• 3

  Run model inference as a separate process using stdio transport to isolate dependencies and credentials.

Description

This repository provides a Node.js MCP server that proxies requests to the OpenAI API using an API key and configurable model (e.g., gpt-4o). It is intended to be added to an MCP host's settings (mcp_settings.json) and launched as a child process (stdio transport), enabling agents to use OpenAI models without embedding the API logic. Configuration is handled via command-line args and environment variables (OPENAI_MCP_API_KEY, OPENAI_MCP_MODEL), making it easy to run locally or integrate into existing MCP setups.