A Model Context Protocol (MCP) server that provides seamless access to OpenAI's o3, o3-mini, and o3-pro reasoning models. This server enables Claude Desktop, Claude Code, and other MCP clients to interact with OpenAI's latest reasoning models through a standardized interface.
The easiest way to use the o3 MCP server is with npx:
npx o3-mcp
This will automatically download and run the latest version without any installation. You'll need to set your OpenAI API key as an environment variable:
export OPENAI_API_KEY="your-openai-api-key-here"
npx o3-mcp
npm install -g o3-mcp
export OPENAI_API_KEY="your-openai-api-key-here"
o3-mcp
git clone https://github.com/GitMaxd/o3-mcp.git
cd o3-mcp
npm install
# Create a .env file
echo "OPENAI_API_KEY=your-openai-api-key-here" > .env
# Edit .env and add your actual OpenAI API key
Add the following to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"o3": {
"command": "npx",
"args": ["o3-mcp"],
"env": {
"OPENAI_API_KEY": "your-openai-api-key-here"
}
}
}
}
{
"mcpServers": {
"o3": {
"command": "node",
"args": ["/absolute/path/to/o3-mcp/index.js"],
"cwd": "/absolute/path/to/o3-mcp"
}
}
}
Important: The cwd parameter is required when using a .env file. It tells the server where to find the .env file.
Note: You may need to use the full path to Node.js if it's not in your system PATH:
{
"mcpServers": {
"o3": {
"command": "/Users/username/.nvm/versions/node/v22.14.0/bin/node",
"args": ["/Users/username/o3-mcp/index.js"],
"cwd": "/Users/username/o3-mcp"
}
}
}
Option 1: Using .env file (Recommended for development)
cwd parameter in configuration.env fileOption 2: Direct configuration (Recommended for simplicity)
{
"mcpServers": {
"o3": {
"command": "node",
"args": ["/absolute/path/to/o3-mcp/index.js"],
"env": {
"OPENAI_API_KEY": "your-api-key-here"
}
}
}
}
With this approach, you don't need the cwd parameter or .env file.
Claude Code uses the same configuration file as Claude Desktop. See the Claude Desktop section above.
For Cline, Continue, or other MCP-compatible clients, refer to their documentation for adding MCP servers. The command remains the same:
node /absolute/path/to/o3-mcp/index.js
Simple, single-turn interaction with o3 models.
Parameters:
prompt (required): Your question or promptmodel (optional): "o3", "o3-mini", or "o3-pro" (default: "o3")developer_message (optional): System context (default: "You are a helpful assistant.")Multi-turn conversations with context preservation.
Parameters:
messages (required): Array of message objects with role and contentmodel (optional): "o3", "o3-mini", or "o3-pro" (default: "o3")max_tokens (optional): Maximum response tokens (default: 1000)Get a quick fact:
Use o3 to tell me an interesting fact about black holes.
Code explanation:
Ask o3 to explain how async/await works in JavaScript.
Advanced analysis with o3-pro:
Ask o3-pro to analyze the computational complexity of quantum algorithms for factoring large numbers.
Algorithm Analysis:
I have this sorting algorithm. Can you ask o3 to analyze its time complexity and suggest optimizations?
[paste your code]
Architecture Review:
Create a dialogue with o3 about the pros and cons of microservices vs monolithic architecture for a startup.
Let's have o3 help us implement a binary search tree. First, ask it for the basic structure, then we'll iterate on adding balancing logic.
I'm going to show you my React component. Can you have o3 review it for performance issues and best practices?
Create a multi-turn conversation with o3 where it explains quantum computing concepts using programming analogies.
My recursive function is causing a stack overflow. Let's debug it together with o3's help.
I need to design a REST API for a social media app. Can you discuss with o3 about best practices for endpoint design, authentication, and rate limiting?
Create a dialogue with o3 about implementing a real-time collaborative editor:
1. First, ask about the architecture
2. Then dive into conflict resolution strategies
3. Finally, discuss scaling considerations
The server includes comprehensive error handling:
"MCP Server Failed to Connect"
.env file, you MUST include "cwd": "/path/to/openai-o3-tool"node isn't in PATHenv object for simplicity"API Key Not Found"
.env file because it's running from wrong directorycwd parameter or include API key directly in config"No output from o3_chat"
"Model not found"
The server logs detailed information to stderr. To see logs when running standalone:
node index.js 2> debug.log
o3-mini for rapid prototyping and testingmax_tokens limits to control costs.env file or API keysContributions are welcome! Please feel free to submit a Pull Request.
git checkout -b feature/AmazingFeature)git commit -m 'Add some AmazingFeature')git push origin feature/AmazingFeature)This project is licensed under the MIT License - see the LICENSE file for details.
Created by @GitMaxd
FAQs
MCP server for OpenAI o3 model integration
The npm package o3-mcp receives a total of 13 weekly downloads. As such, o3-mcp popularity was classified as not popular.
We found that o3-mcp demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 1 open source maintainer collaborating on the project.
Security News
The PHP Foundation says Chainguard broke a security embargo ahead of critical RCE patches, as leaks disrupted the coordinated release.
Security News
PHP releases emergency patches for two critical RCE vulnerabilities as agencies urge organizations to take unpatched sites offline.
Threat Research
/Security News
A supply chain attack compromised LiteLLM 1.107.0 on PyPI and Docker, spreading malware that steals cloud credentials, developer secrets, and AI provider keys.