OpenAI announced new features for its software engineering agent Codex at Dev Day, held on September 29, 2026. The updates aim to make Codex usable beyond developers’ computers and turn cloud tasks into persistent workspaces.
- Reusable cloud development environments will offer access from different devices, faster task initiation, and shared workspaces that include permissions with approved settings for teams.
- The revamped Codex CLI will gain the ability to start and direct tasks by voice. The new “/agents” view will make it easier to delegate and monitor multiple tasks; prompt editing, resuming sessions, and worktree use will also be improved. A more readable terminal interface will also be provided.
- The new code review experience in the ChatGPT desktop app will allow users to read change summaries, review code, and ask Codex about potential issues before GitHub pull requests or GitLab merge requests. Automated reviews will be able to perform an initial assessment even when users step away from their computers.
- Codex Security Cloud will scan GitHub repositories on demand, on a schedule, or when new commits arrive; review findings, separate iterations, and prepare fixes in the cloud. The tools will provide access to models covered by Daybreak Blue without requiring a separate application.
- OpenAI announced the Decisions API, which answers user-defined questions with predefined responses via Luna for real-time decision-making. The updated Agents API will support computer use and the creation of OpenAI agents that run entirely on AWS with Amazon Bedrock Managed Agents.
Why it matters
This package aims to transform Codex from an assistant running on a single computer into part of a development workflow that can be sustained across different devices and teams. Persistent cloud workspaces, shared settings, and task tracking could particularly affect the coordination of teams working on the same codebase; however, how permissions will be managed and how responsibilities will be shared within teams remain open implementation questions. While automating code reviews and security scans could reduce the time developers spend on initial checks, how fixes prepared in the cloud will undergo human oversight is important. API innovations expand the option of bringing these capabilities to work environments outside OpenAI and to AWS infrastructure; the scope and terms of use are not detailed in the text.
Background
OpenAI is not a new name in the FikirPilot archive: we have published 60 news articles mentioning this name in the last 90 days; the most recent was dated 29 September 2026.
Term: agent
An artificial intelligence agent is software that calls tools and carries out multi-step tasks to achieve a goal, rather than generating a single response.