OpenAI has published new configuration options for Codex on its OpenAI Academy documentation, giving developers more control over how the AI coding assistant behaves in their projects.
According to the documentation, the update introduces three categories of settings:
- Personalization: Codex can adapt to a user's or team's coding style and recurring conventions, aiming for more consistent, context-aware suggestions.
- Detail Level: Developers can adjust how verbose or detailed the generated code is, from high-level boilerplate to fully implemented functions.
- Permissions: Fine-grained controls let teams restrict which files or directories Codex can access within a codebase.
OpenAI frames these settings as a way to make Codex behave less like a generic autocomplete tool and more like an assistant tailored to a specific project's conventions and security requirements. The permissions controls in particular address a common concern for teams wary of giving AI tools broad access to a codebase.
The documentation does not include performance benchmarks or cost figures for the new settings; developers evaluating the impact on token usage or API costs will need to test within their own workflows.