OpenAI upgrades Codex with reusable cloud workspaces across devices

OpenAI has expanded Codex with reusable cloud development environments, making its software engineering agent usable across laptops, phones and the cloud. The update matters because it turns Codex from a remote task runner into a more persistent collaboration tool for teams that want faster starts, shared settings and continuous work on code.

The new capabilities were unveiled at OpenAI’s Dev Day on Tuesday alongside several other Codex changes, including a refreshed command-line interface, a new code review workflow, security tooling and API updates. Together, the releases point to OpenAI’s effort to make Codex a broader development platform rather than a single-purpose coding assistant.

What changed in Codex?

OpenAI is moving Codex beyond isolated cloud jobs and toward reusable workspaces that can follow developers across devices. That means a task can be started on one machine, resumed elsewhere and managed in a shared environment with pre-approved settings and permissions.

Previously, Codex could already launch cloud tasks to handle remote work. The new version goes further by making those environments more persistent and configurable, which should reduce setup time and make it easier for teams to standardize how they work.

Why does that matter?

It matters because modern software teams rarely work from one place or one device. A developer might begin debugging at a desk, check progress from a phone and later continue in the cloud. Reusable environments help keep context intact, lower friction and make collaborative work easier to coordinate.

OpenAI said the new environments are designed to help tasks launch more quickly and create a common workspace for approved tools, credentials and permissions. In practice, that could make Codex more attractive for distributed teams and for companies that want tighter control over how AI agents access internal code.

How is OpenAI changing the Codex workflow?

OpenAI is adding several workflow features intended to make Codex easier to steer, review and resume. The company is also trying to improve the experience for long-running sessions, where coding agents can accumulate a lot of context and require more careful oversight.

• Updated Codex CLI: Developers can now start and direct tasks using voice input.
• New /agents view: This interface is built to help people delegate work and monitor multiple tasks at once.
• Session improvements: Prompt editing, session resumption and worktree handling have been refined.
• Cleaner terminal UI: A redesigned text interface aims to make longer sessions easier to follow.

For developers who already rely on terminal-based tools, these changes could be especially useful. Voice control may speed up simple instructions, while the new task view should make it easier to keep track of several agent-driven jobs at the same time.

How does Codex fit into code review?

OpenAI is now bringing Codex into a new code review flow inside the ChatGPT desktop app. That feature allows users to inspect summaries, explore code changes and ask questions about possible issues before they respond to pull requests or merge requests in GitHub and GitLab.

The review workflow also includes automatic code checks, allowing Codex to complete an initial pass while developers are away from their machines. That could help teams catch obvious problems earlier, shorten review cycles and reduce the amount of manual triage needed before human review begins.

OpenAI is positioning Codex as a tool that can assist with both active development and asynchronous review, letting teams hand off routine checks while still keeping final decisions with humans.

That combination is important. Coding agents are increasingly expected not just to generate snippets, but to participate in the full software lifecycle: drafting code, reviewing changes, spotting bugs and helping maintain projects over time.

What is Codex Security Cloud?

Codex Security Cloud is OpenAI’s new security-focused toolset for scanning codebases and helping teams fix issues in the cloud. The feature is designed to examine full GitHub repositories either on demand or on a schedule, and it can also run when new commits land.

Once a scan is underway, Codex can investigate findings, remove duplicate alerts and prepare fixes even if the developer is offline and the laptop is closed. That gives teams a way to keep security checks moving continuously instead of waiting for a person to trigger each review.

OpenAI said the toolset includes access to models from its Daybreak Blue cybersecurity initiative, and developers do not need to submit a separate application to use them. That lowers the barrier to entry for teams that want to experiment with AI-assisted security workflows.

Why this security push matters

Security tooling has become one of the most competitive areas for AI coding products because companies want faster detection without sacrificing control. By tying Codex to repository scanning and fix preparation, OpenAI is clearly trying to make the agent useful for more than code generation alone.

For businesses, the appeal is straightforward: fewer delayed scans, faster triage and a better chance of catching issues before they move deeper into production systems.

What are the new API updates?

OpenAI also used Dev Day to update its APIs, extending how developers can build agentic applications around Codex and related tooling. The most notable additions include a new decision-making API and a more capable Agents API.

What is the Decisions API?

The Decisions API is OpenAI’s new real-time decision-making tool. It uses Luna to respond to a defined set of questions with predetermined answers, which suggests a design aimed at structured workflows rather than open-ended conversation.

That makes the feature potentially useful for business processes where consistency matters more than creative generation, such as routing requests, checking conditions or triggering specific actions.

What changed in the Agents API?

OpenAI said its updated Agents API now supports computer use and can work with Amazon Bedrock Managed Agents to build OpenAI agents that run fully on AWS. That expands the range of environments where developers can deploy AI systems and may appeal to companies already invested in Amazon’s cloud stack.

Support for computer use also suggests OpenAI wants agents to interact more directly with digital interfaces, not just produce text or code. That capability has become increasingly central to the next generation of AI tools.

How does this fit into OpenAI’s broader strategy?

OpenAI’s Codex updates fit a larger trend in the AI industry: moving from chat-based assistance to integrated agent platforms that can work across devices, handle tasks independently and plug into real enterprise workflows.

Instead of treating coding as a one-off prompt-and-response problem, OpenAI is building around the full lifecycle of software development. That includes writing code, reviewing it, checking security, resuming work later and coordinating with other agents or systems.

This approach also helps OpenAI compete more directly with other developer-focused AI products that emphasize automation, persistence and collaboration. The more Codex can function like a team environment rather than a single assistant, the more deeply it can be embedded into day-to-day engineering operations.

Why developers should pay attention

Developers and engineering managers should pay attention because the update changes how AI assistance can be used in practical workflows. The biggest benefits are likely to show up in teams that juggle multiple repos, review cycles and security checks.

1. Faster setup: Reusable environments reduce repeated configuration work.
2. Better continuity: Tasks can move more smoothly between devices and sessions.
3. Shared controls: Approved settings and permissions make team use more predictable.
4. More automation: Review and security features can run while developers are offline.
5. Broader deployment: API updates extend Codex-style agents into more cloud architectures.

For smaller teams, the most immediate value may be convenience. For larger organizations, the more significant change may be governance: the ability to define how AI agents operate inside controlled environments and to inspect their work before it reaches production.

Timeline of the Codex rollout

OpenAI’s announcement did not come as a single isolated feature drop. It was part of a wider Dev Day package that suggests a staged expansion of Codex’s role across development, review and security.

What it means for the market

OpenAI’s latest move signals that the competitive front in AI is no longer just about model capability. It is about how well a company can package those models into usable systems for developers, security teams and enterprise customers.

Persistent environments, voice-enabled CLI tools, automated review and cloud security scanning are all part of the same message: AI should not just answer questions, it should keep working after the question is over. That is the direction in which many software teams are now heading.

As AI agents become more capable, the challenge shifts from whether they can perform tasks to whether organizations can trust them to do so reliably. OpenAI’s Codex release is a clear attempt to address both the speed question and the trust question at once.

Bottom line

OpenAI’s Codex update is more than a convenience tweak. By introducing reusable cloud environments, deeper review tools, automated security scanning and new APIs, the company is pushing its coding agent toward a more complete developer platform that can work across devices and stay active even when users are not.

For teams already experimenting with AI-assisted software engineering, the update offers a clearer path from chat-style assistance to an always-available workflow layer. For the broader market, it is another sign that agentic software is becoming a central battleground in enterprise AI.