Codex gets reusable cloud environments, voice control in the CLI, and repository security scans
OpenAI turned Codex cloud tasks into published environments that a team can reuse, with each task in its own workspace and work continuing while your laptop sleeps. The CLI takes spoken instructions and shows parallel tasks in a new view. Codex Security Cloud scans whole GitHub repositories and prepares fixes in the cloud. The Agents API now drives a computer, and a Bedrock version runs the same agents inside AWS.
OpenAI announced a set of Codex and API changes at DevDay on September 29, 2026. The common thread is that work moves off the developer's machine: environments live in the cloud, reviews and security scans run there without the laptop open, and the same agent stack becomes available inside AWS.
Cloud environments a team can publish and reuse
Codex could already run tasks remotely, but each one started from scratch. Now a cloud environment is created once and published: Codex inspects the chosen GitHub repositories, installs dependencies and tools, tests the workflow, and asks for anything it is missing. After that, every task starts from the published environment, gets its own workspace, and keeps working while the developer's computer is asleep.
The environment holds the access a project needs. Network secrets cover allowed HTTPS services, and the environment limits which network destinations Codex can reach. Tasks can be started from the web, the desktop app, the phone, or the terminal with codex cloud. The feature is available on Plus, Pro, Business, Healthcare, Education, and Enterprise plans.
A CLI you can talk to
The refreshed Codex CLI lets you start and steer tasks by voice. A new /agents view lists the tasks running in parallel so delegated work can be tracked from one place. OpenAI also lists smaller changes: editing prompts, resuming sessions, using worktrees, and a cleaner terminal interface for long sessions. The CLI changes apply to all plans.
Code review in the desktop app
The ChatGPT desktop app gains a code review surface. It shows a pull request's description, changed files, comments, and checks, and lets you ask Codex about a change before posting feedback to GitHub or GitLab. GitHub review is generally available; GitLab merge requests are in preview. With automatic reviews enabled, Codex takes a first pass in the cloud, and that path does not require a cloud environment of its own.
Security scans that run with the laptop closed
Codex Security Cloud is a plugin that scans entire GitHub repositories on demand or on a schedule and keeps checking new commits. Codex investigates each finding, removes duplicates, and prepares fixes in the cloud. A finding that offers a fix produces a proposed patch and, after review, a draft pull request. Monitoring settings choose the cloud environment, how many days of history to review, and whether monitoring is paused.
The plugin includes access to the models offered through OpenAI's Daybreak Blue defensive security program without a separate application. It is available to Pro, Business, Enterprise, and Edu users on desktop and web.
Agents API with computer use, and the same agents on AWS
The Agents API now supports computer use, so an agent built on it can operate software to complete a task. OpenAI says it also brings the multi-agent setup, tool search, tool calling, and context compaction from Codex into third-party applications. It is available through the API and in Codex and ChatGPT Work on the Pro 500 and Enterprise plans.
A new Decisions API, in limited preview, points the Luna model at a fixed set of questions with predefined answers, from text or images, to classify content, route requests, or pick an agent's next action. OpenAI plans a broad release in the coming days.
OpenAI also worked with Amazon on Bedrock Managed Agents, which takes the Agents API capabilities and runs them natively in AWS. Amazon says the service is built on the OpenAI harness, that every agent has its own identity and logs every action, that inference runs on Amazon Bedrock, and that data does not leave AWS. It is in limited preview.
„Codex investigates findings, removes duplicates and prepares fixes in the cloud, even with your laptop closed.“
Sources
Related
Apple will add controls to Full Disk Access on macOS, citing risks from AI agents
Apple said on October 2, 2026, that it will add controls to Full Disk Access, the macOS permission that lets an app read nearly everything on a Mac. The company says some developers use the permission in ways that expose files, mail, messages, and browsing history without users fully understanding it. Apple expects the risk to grow as AI agents become more capable and autonomous. The post gives no date, no macOS version, and no description of the new controls.
Appleverified
AWS previews Well-Architected Agent, limited to Business+ Support plans and above
AWS announced a preview of AWS Well-Architected Agent on October 1, 2026. The service scans your AWS accounts on a schedule and returns recommendations for cost, security, resilience, and performance, ranked against business goals that you write yourself. Remediation scripts come attached, and the agent also reviews infrastructure-as-code templates on demand. Access requires an AWS Support plan at the Business+ tier or higher, and agent profiles are hosted in three US Regions.
AWSverified
Claude Code 2.1.287 adds Mods and makes a 1M context window the default on cloud platforms
Anthropic released Claude Code 2.1.287 on October 1, 2026. The release adds Claude Mods, which let plugins modify deeper behavior, and a built-in mod in which a side agent flags things you might miss. Opus 4.7 and later and Fable now use a 1M context window by default on Bedrock, Vertex, Foundry, and the Claude apps gateway. Some MCP servers may stop connecting until you add one line to their configuration, and several shell writes that used to run now wait for approval.
Claude Codeverified