OpenAI announced a slate of Codex and API updates at DevDay in San Francisco on September 29, 2026. The developer-focused changes include reusable cloud environments, a voice-enabled CLI, code review and repository security tools, plus Computer Use in the Agents API, the Decisions API and GPT-6 Astra Ultrafast. Access depends on the feature and plan.
Codex gets cloud environments, voice controls and code review
Codex cloud environments let teams reuse a development setup with shared settings and permissions. OpenAI said tasks can be started from a computer or phone and run in the cloud.
Codex cloud environments
The Codex cloud-projects interface presents projects and a Tasks area in its Cloud view.
Refreshed Codex CLI
The updated Codex CLI supports voice-started and voice-steered tasks. It also adds an /agents view for delegating and tracking multiple tasks, along with prompt editing, session resumption, worktrees and a cleaner terminal interface. OpenAI listed the CLI for all plans.
Code Review and repository security
Code Review presents summaries and code diffs, and lets users ask Codex about potential issues. Automatic first-pass reviews can run in the cloud; the feature supports GitHub pull requests and GitLab merge requests. OpenAI listed Code Review for all plans.
Codex Security Cloud scans GitHub repositories on demand or on a schedule, checks new commits, investigates and deduplicates findings, and prepares fixes in the cloud. OpenAI listed it for Pro, Business, Enterprise and Edu users on desktop and web. The offering includes Daybreak Blue models without a separate Daybreak application.
Agents API Computer Use and the Decisions API
OpenAI added Computer Use to the Agents API, enabling agents to interact with software. The API also includes Codex multi-agent capabilities, tool search and calling, and context compaction. OpenAI listed access through the API, as well as through Codex and ChatGPT Work for Pro 500 and Enterprise.
The Decisions API applies Luna’s intelligence to a developer-defined question with a finite set of answers. It accepts text or image context for tasks such as classification, routing a request or choosing an agent’s next action. OpenAI announced it in limited preview on September 29 and said a broader release was planned for the coming days.
| Offering | What it does | Access OpenAI listed at announcement |
| Codex cloud environments | Reusable team setup; tasks can run from a computer, phone or cloud | Plus, Pro, Business, Healthcare, Education and Enterprise |
| Codex CLI | Voice controls, task management and terminal updates | All plans |
| Code Review | Summaries, diffs, questions about potential issues and cloud-based first-pass reviews | All plans |
| Codex Security Cloud | GitHub repository scans, finding investigation and cloud-prepared fixes | Pro, Business, Enterprise and Edu on desktop and web |
| Agents API Computer Use | Lets agents interact with software; includes other Codex agent and tool capabilities | API; also Codex and ChatGPT Work for Pro 500 and Enterprise |
| Decisions API | Makes a choice from a finite, developer-defined set using text or image context | Limited preview; broader release planned in the coming days |
| GPT-6 Astra Ultrafast | Faster token generation in Codex and the API | API; ChatGPT Work and Codex for Pro 500 and Enterprise |
GPT-6 Astra Ultrafast: OpenAI’s speed claims
OpenAI says GPT-6 Astra Ultrafast can generate up to 300 tokens per second in Codex, with up to 8× faster generation there. For the API, the company claims up to 6× speed. These are OpenAI’s stated figures.
OpenAI’s DevDay presentation includes a demonstration of Ultrafast.
Pro 500’s US terms
OpenAI announced Pro 500 in the US at $500 per month. The plan includes Ultrafast and 25 times the ChatGPT Plus usage allowance.
