OpenAI Codex Update Adds Computer Control for 3M Weekly Developers
OpenAI has released a significant overhaul to its Codex coding assistant, expanding the tool's capabilities well beyond traditional code generation to include direct computer control, persistent memory, and automated scheduling—features that position it as an always-on development partner rather than a simple autocomplete tool.
The update, rolling out April 16, 2026, serves the more than 3 million developers who use Codex weekly. The headline feature: background computer use that lets multiple AI agents operate your Mac simultaneously without disrupting your own work.
What's Actually New
Codex can now see your screen, click, and type with its own cursor across any application—not just code editors. For developers, this means the agent can test frontend changes, interact with apps lacking APIs, or iterate on designs without constant manual intervention.
The addition of an in-app browser lets developers comment directly on web pages to guide the agent, particularly useful for frontend work and game development. OpenAI plans to expand browser capabilities beyond localhost applications over time.
Image generation via gpt-image-1.5 now integrates directly into workflows, allowing Codex to create and iterate on visuals for mockups, product concepts, and game assets alongside code.
More than 90 new plugins expand the tool's reach into enterprise workflows: Atlassian Rovo for JIRA management, CircleCI, GitLab Issues, Microsoft Suite, Neon by Databricks, and others. These combine with MCP servers to let Codex gather context and take action across disparate tools.
Persistent Memory Changes the Game
Perhaps the most significant shift is Codex's new memory system. The agent now retains context from previous sessions—personal preferences, corrections, and hard-won information. This means less time re-explaining project conventions and more time getting actual work done.
Codex can also schedule future work for itself, waking automatically to continue long-running tasks across days or weeks. Teams are using these automations to land open pull requests, follow up on tasks, and monitor conversations across Slack, Gmail, and Notion.
The tool now proactively suggests work based on your projects, connected plugins, and memory. It can identify open Google Docs comments requiring attention, pull context from multiple sources, and present prioritized action lists.
Availability and Limitations
The update rolls out immediately to Codex desktop users signed in with ChatGPT. Computer use starts on macOS only. Enterprise, Edu, and EU/UK users will receive personalization features including memory and context-aware suggestions in a subsequent rollout.
For crypto developers specifically, the expanded plugin ecosystem and computer use capabilities could streamline smart contract testing, documentation management, and cross-tool coordination—though the real test will be whether the memory system can reliably retain blockchain-specific context across sessions.
OpenAI frames this release as part of its AGI mission to narrow the gap between imagination and execution. One year after Codex launched, usage patterns have expanded from code writing to system understanding, debugging, team coordination, and long-running project management. Whether this update delivers on that expanded promise depends entirely on execution—and how well those 3 million developers actually integrate these features into production workflows.