Skip to main content

Enterprise admin getting started guide for Codex

Updated over 3 weeks ago

Codex in ChatGPT

Codex is OpenAI’s cloud-based software engineering agent—purpose-built to support professional developers by securely connecting to your codebase and automating tasks like bug fixing, feature development, test generation, and more. It runs tasks in parallel in the cloud, so engineers stay focused while Codex handles the rest.

This guide is tailored for ChatGPT Enterprise Admins looking to set up Codex. If you’re a developer, check out our docs.

Note: To set up Codex as an admin, you must have GitHub access to the repositories

commonly used across your organization. If you don’t have the necessary access, you’ll need to collaborate with someone on your Engineering team who does.

Enterprise-ready security

Codex automatically supports all ChatGPT Enterprise security features, including:

  • Data Retention & Residency compliance

  • Inclusion in Compliance APIs

  • No training on your data—ever

Prerequisites

Codex currently supports GitHub (cloud-hosted) as its source code management system. To get started, you’ll likely need Admin access to your organization’s GitHub environment.

Setup

Enable Codex in workspace settings

Start by toggling Codex ON for your organization.

You’ll find the toggle under Manage Workspace > Settings. Once enabled, users can access Codex directly from the left-hand navigation panel in ChatGPT.

⚠️ Please note: After you toggle Codex to ON in your Enterprise workspace settings, it will take around 10 mins for the Codex UI element to populate in ChatGPT.

Setup your your first environment

  1. Navigate to Codex and click Get Started to begin onboarding.

  2. Click Connect to GitHub to start installation of the ChatGPT GitHub Connector if you have not already connected to GitHub with ChatGPT.

    • Authorize the ChatGPT Connector for your user

    • Choose your installation target for the ChatGPT Connector (typically your main organization)

    • Authorize the repositories you’d like to enable to connect to Codex (may require a GitHub admin to approve).

  3. Create your first environment by selecting the repository most relevant to your developers. Don’t worry, you can always add more later. Then click Create Environment

  4. Codex will suggest starter tasks (e.g. writing tests, fixing bugs, exploring code) that can run concurrently; click Start Tasks button to kick them off.

You have now created your first environment. Individuals who connect to GitHub will now be able to create tasks using this environment and users who are authorized for the relevant repository will have the ability to push pull requests generated from their tasks.

Connect additional GitHub repositories with Codex

  1. Click the Environments button or open the environment selector and click Manage Environments.

  2. Click the Create Environment button

  3. Select the environment you’d like to connect to this environment

  4. Give the environment an easy to grok name and description.

  5. Select the environment visibility

  6. Click the Create Environment button

Note: We strongly recommend that you do not perform any advanced configuration until you have been using Codex with the basic environment configuration for at least a month. After you understand Codex’s out-of-the-box performance and are ready to make optimizations you can use the advanced configuration to make adjustments to help potentially boost performance. View our docs to learn more.

Help an end user get started with Codex

The following are instructions you can share with your end users on how to get started using Codex:

  1. Navigate to Codex in the left-hand panel.

  2. Click the Connect to GitHub button inside of the prompt composer if not already connected

    • Authenticate into GitHub

  3. The user is now able to use publicly shared environments or create their own environment.

  4. Try getting started with a task using both Ask and Code mode, here is something you can try:

    • Ask mode: What bugs do I have?

    • Code mode: In {project_name}, create unit tests for code with zero coverage, following our existing test pattern.

Did this answer your question?