OpenAI provides several channels for reporting bugs and inaccuracies in Codex, with different approaches depending on whether you’re using the web-based ChatGPT integration or the CLI tool. For issues with the web-based Codex agent accessed through ChatGPT, you can report problems directly through the ChatGPT interface using the feedback mechanisms built into the platform. This includes thumbs up/down ratings on specific responses and detailed feedback forms that allow you to describe issues with generated code, unexpected behavior, or feature requests. OpenAI actively monitors these feedback channels and uses the information to improve the system’s performance and address common issues that users encounter.
For the open-source Codex CLI tool, bug reporting follows standard GitHub practices since the project is hosted on GitHub under the OpenAI organization. You can file issues directly in the repository by creating detailed bug reports that include information about your operating system, CLI version, the specific commands that caused problems, and any error messages or unexpected behavior you encountered. The development team encourages users to provide reproducible test cases and clear descriptions of the expected versus actual behavior. You can also contribute fixes through pull requests if you identify and resolve issues yourself, as the CLI tool welcomes community contributions for bug fixes and improvements.
When reporting any issues with Codex, it’s helpful to provide specific context about the task you were trying to accomplish, the programming language and frameworks involved, and the exact input that led to problematic output. For security vulnerabilities or issues that could affect user safety, OpenAI has established responsible disclosure procedures that allow you to report sensitive issues through secure channels. Additionally, the OpenAI community forums and Discord channels provide spaces for discussing issues with other users and getting help with problems that may not be bugs but rather usage questions or configuration issues. The development team actively participates in these community discussions and often provides guidance on best practices and workarounds for known limitations while working on longer-term fixes.