---
title: "OpenAI Codex at UConn"
canonical: "https://kb.uconn.edu/space/AI/29080944706/OpenAI+Codex+at+UConn"
format: markdown
---
Codex is OpenAI's coding agent. It can help users understand, write, modify, test, and review software by working from natural-language instructions and, where enabled, interacting with code repositories and development environments.

## In this article

> Macro (toc)

## Who can use Codex?

- **Faculty and staff:** Eligible UConn employees with access to the university-licensed ChatGPT service may use Codex for approved university work.
- **Student workers:** Eligible student employees may use Codex when acting in their university work role and when their account and project are supported.

Access depends on UConn licensing, account eligibility, product availability, and any administrator controls applied to the workspace.

## How to Purchase Access to Codex

You’ll need to purchase** **a UConn-licensed ChatGPT account at [software.uconn.edu](https://software.uconn.edu/software/chatgpt/).

[Codex usage consumes credits](https://help.openai.com/en/articles/11481834-chatgpt-rate-card-business-enterpriseedu-credit-based-pricing#chatgpt-work-and-codex). Any significant or sustained use likely will require membership in the **Power ChatGPT tier**. The Base Tier may be sufficient for occasional, limited experimentation, but it is not expected to provide sustained coding-agent workflows, large repositories, repeated test runs, or other high-compute use. Power Tier usage is subject to usage-based billing and applicable monthly limits. See the [ChatGPT at UConn](https://uconn.atlassian.net/wiki/spaces/AI/pages/29081010275) article for more information about the two tiers.

> ℹ️ Codex access does not make the OpenAI API free; API usage is a separate service with separate billing.

## How to Access Codex

1. Sign in to the university-licensed ChatGPT service with your UConn account.
2. Open the Codex experience from the ChatGPT interface, the Codex application, or the VS Code plugin.
3. Confirm that the intended repository, workspace, tools, and permissions are selected.

## What can Codex do?

- Explain unfamiliar code, project structure, and dependencies.
- Draft new code, scripts, tests, configuration, and technical documentation.
- Suggest bug fixes, refactoring, and performance improvements.
- Review changes for correctness, maintainability, security, and potential regressions.
- Help troubleshoot build failures, test failures, and development-tool errors.
- Automate repetitive development tasks when the user has reviewed and authorized the proposed changes.

Codex generates suggestions and may make changes within an authorized workspace. Users remain responsible for reviewing code, running appropriate tests, validating dependencies and licenses, and approving changes before they are committed, deployed, or used in production.

## Credit usage and recommended practices

Credits may be consumed by model requests and advanced coding-agent activities. Large repositories, long conversations, repeated retries, extensive file analysis, test execution, and multi-step tasks may consume credits more quickly than short coding questions.

- Use Codex for tasks that benefit from codebase context or multi-step reasoning.
- Keep requests focused and provide the relevant files, commands, and expected outcome.
- Use smaller scopes and incremental changes instead of asking for a broad rewrite immediately.
- Review the planned changes before allowing edits, commands, or tests to run.
- Avoid repeatedly retrying the same request without changing the instructions or investigating the cause.
- Monitor credit consumption in ChatGPT under **Settings** and **Usage** or **Usage & billing**.

## Security and privacy

For institutional work, use Codex only through a UConn-supported account and sign in with your official UConn credentials. UConn's agreement with OpenAI states that prompts, responses, and data from the university-licensed service are not used to train OpenAI's models when users are signed in through the university workspace.

Do not provide Codex with secrets, passwords, private keys, access tokens, or other credentials. Do not place protected or confidential data in a repository or workspace unless the use is permitted by UConn policy, the applicable data owner, and the service agreement. Review repository permissions and connected tools before granting access.

AI-generated code can contain security vulnerabilities, incorrect assumptions, licensing concerns, or destructive commands. Treat Codex output as untrusted until it has been reviewed and tested by a qualified person.

## When do we recommend Codex?

| Need | Recommendation |
| --- | --- |
| Explain a function, error message, or small code sample | **Codex may be appropriate.** Review the explanation and verify it against the source code and documentation. |
| Develop or refactor a substantial application | **Use Codex with Power Tier access.** Establish code review, testing, and change-control practices before work begins. |
| Work with sensitive data, secrets, or production credentials | **Do not use Codex without review and approval.** Remove secrets and confirm that the proposed data handling is permitted. |
| Deploy code or make irreversible changes | **Do not rely on Codex alone.** Require qualified human review, testing, backups, and the appropriate release process. |
| Build an application that calls an OpenAI model | **Evaluate the OpenAI API separately.** API access, provisioning, security review, and billing are separate from ChatGPT and Codex. |