Package Exports

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (openai-plugins) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

openai-mcp

A TypeScript library that provides an OpenAI-compatible client for the Model Context Protocol (MCP).

Installation

npm install openai-mcp

Features

OpenAI API compatibility - works as a drop-in replacement for the OpenAI client
Connects to local or remote Model Context Protocol servers
Supports tool use and function calling
Rate limiting and retry logic built in
Configurable logging
TypeScript type definitions included

Usage

import { OpenAI } from 'openai-mcp';

// Create an OpenAI-compatible client connected to an MCP server
const openai = new OpenAI({
  mcp: {
    // MCP server URL(s) to connect to
    serverUrls: ['http://localhost:3000/mcp'],
    
    // Optional: set log level (debug, info, warn, error)
    logLevel: 'info',
    
    // Additional configuration options
    // modelName: 'gpt-4',             // Default model to use
    // disconnectAfterUse: true,       // Auto-disconnect after use
    // maxToolCalls: 15,               // Max number of tool calls per conversation
    // toolTimeoutSec: 60,             // Timeout for tool calls
  }
});

// Use the client like a standard OpenAI client
const response = await openai.chat.completions.create({
  model: 'gpt-4',
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Hello, how are you today?' }
  ]
});

console.log(response.choices[0].message.content);

Logging Configuration

import { setMcpLogLevel } from 'openai-mcp';

// Set log level to one of: 'debug', 'info', 'warn', 'error'
setMcpLogLevel('info');

Environment Variables

The library also supports configuration through environment variables:

# MCP Server URL(s) - comma separated for multiple servers
MCP_SERVER_URL=http://localhost:3000/mcp

# API Keys for different model providers
OPENAI_API_KEY=your-openai-api-key
ANTHROPIC_API_KEY=your-anthropic-api-key
GEMINI_API_KEY=your-gemini-api-key

Multi-Model Support

The library supports routing requests to different model providers based on the model name:

import { OpenAI } from 'openai-mcp';

const openai = new OpenAI();

// Uses OpenAI API
const gpt4Response = await openai.chat.completions.create({
  model: 'gpt-4',
  messages: [{ role: 'user', content: 'Hello GPT-4' }]
});

// Uses Anthropic API
const claudeResponse = await openai.chat.completions.create({
  model: 'claude-3',
  messages: [{ role: 'user', content: 'Hello Claude' }]
});

// Uses Google Gemini API
const geminiResponse = await openai.chat.completions.create({
  model: 'gemini-pro',
  messages: [{ role: 'user', content: 'Hello Gemini' }]
});

Examples

The examples/ directory contains various usage examples:

Basic Usage: Simple chat completion request
Streaming: Stream responses token by token
Multi-Model: Use OpenAI, Anthropic, and Gemini models
Tools Usage: Function/tool calling with MCP
Custom Logging: Configure and use the logging system

See the Examples README for more details on running these examples.

Development

To build the library:

npm run build

To run tests:

npm test