Codex is an agentic AI coding assistant from OpenAI that works directly within your development environment. It can explore and explain codebases, write and modify code, debug problems, run commands and tests, review changes, and help implement new features.
Codex can be used from your terminal through the Codex command-line interface (CLI) and within supported integrated development environments (IDEs). U-M AI Services provides access to Codex models through U-M GPT Toolkit, allowing you to use Codex without managing a separate OpenAI API account.
When configured for U-M GPT Toolkit, Codex runs locally on your computer while model requests are sent through the Toolkit gateway. Features that depend on OpenAI-hosted Codex services may not be available through the Toolkit configuration.
There are costs associated with this service, billed through U-M GPT Toolkit. See our Pricing page for more information.
Who This Is For
Codex via U-M GPT Toolkit is intended for faculty and staff who want an AI coding assistant integrated into their existing development workflow.
Sensitive Data
Check the ITS AI Service entry in the Sensitive Data Guide to IT Services for a list of the data types that are permitted to be used with ITS AI Services.
Before using institutional and research data with third-party AI services, carefully review the ITS Safe Computing guidelines.
How to Access Codex
To get started with Codex, you'll need to create an API key through the U-M GPT Toolkit. Follow the instructions in Create and Retrieve a U-M GPT Toolkit Key to generate your key. Once you have your key, see the setup guides below to install and configure Codex on your machine.
Support & Documentation
For general questions about Codex via U-M GPT Toolkit, see our FAQ page.
For Gateway API support (authentication, endpoint configuration, connectivity, and other gateway setup questions), see U-M GPT Toolkit In Depth.
For guidance on Codex API Keys and setup instructions for Mac, Windows, and Linux, see: