OpenAI Expands Codex Into Full Software Development Assistant
OpenAI has upgraded Codex with computer control, memory, and workflow integrations. The update pushes Codex beyond coding into a full development lifecycle assistant.
OpenAI has released a major update to Codex, significantly expanding its capabilities beyond code generation into a broader software development assistant. The update targets the more than 3 million developers using Codex weekly and reflects a growing push toward AI systems that can manage end-to-end workflows rather than isolated tasks.
The new version introduces “computer use,” allowing Codex to interact directly with a user’s device by seeing screens, clicking, and typing via its own cursor. Multiple AI agents can run in parallel on macOS without interfering with user activity. This enables tasks such as testing applications, iterating on front-end designs, and working with tools that lack APIs.
Codex now also includes an in-app browser, enabling developers to annotate web pages and guide the AI in real time. This feature is aimed at improving workflows in frontend and game development, with plans to expand toward broader browser automation. In parallel, Codex gains image generation capabilities through integration with OpenAI’s image model, allowing developers to create and refine visual assets such as mockups and UI concepts directly within the development process.
Deeper Integration Across Developer Tools
The update introduces more than 90 new plugins, expanding Codex’s ability to connect with commonly used tools and services. These include integrations with platforms such as Jira, GitLab, CircleCI, and Microsoft Office, among others. The plugins combine app integrations and external servers to give Codex more context and execution capabilities across workflows.
Within the Codex app, new features support key development tasks such as reviewing pull requests, addressing code review comments, and managing multiple terminal sessions. Developers can also connect to remote development environments via SSH and access files with rich previews for documents, spreadsheets, and presentations. A new summary pane helps track agent actions, sources, and outputs.
These additions are designed to reduce context switching, allowing developers to move between writing code, reviewing outputs, and collaborating with AI in a single environment.
Automation, Memory, and Long-Running Tasks
Codex is also gaining stronger automation capabilities. Users can now reuse conversation threads to preserve context across sessions and schedule tasks for the future. The system can “wake up” and continue work on long-running processes, spanning days or weeks.
A preview of memory features allows Codex to retain user preferences, corrections, and previously gathered information. This enables more personalized and efficient task execution over time, reducing the need for repeated instructions.
Additionally, Codex can proactively suggest next steps based on project context. For example, it can identify unresolved comments in documents, pull updates from tools like Slack or Notion, and generate a prioritized task list to help users resume work quickly.
Availability and Direction
The update is rolling out to Codex desktop users signed in with ChatGPT, with some features such as memory and personalization expanding later to enterprise and regional users. Computer control capabilities are initially limited to macOS.
The release highlights a broader shift in AI development tools. Codex is evolving from a coding assistant into a system that can coordinate tasks, manage workflows, and assist with decision-making across the software lifecycle. This positions it closer to an autonomous development partner rather than a reactive tool, reflecting how developers are increasingly using AI not just to write code, but to manage complex projects end to end.
Mistral's reported €3 billion round would fund a 1-gigawatt European compute buildout, yet leaves the OpenAI rival far smaller than US and Chinese rivals.
