Writing Team
.png)
OpenAI's Codex desktop app is no longer just a coding assistant. As of this week, it can operate your computer, generate images, remember your preferences, schedule its own future tasks, and work across virtually every app on your machine — while you're using it for something else entirely.
That's the update. It's a significant one.
Codex Desktop App: The Basics
Codex is OpenAI's AI-powered development environment, used by more than 3 million developers weekly. It started as a code-generation tool and has steadily expanded into a full software development lifecycle assistant — writing code, reviewing pull requests, debugging, and coordinating across tools such as Slack, GitHub, Notion, and Gmail.
The desktop app is the primary interface. It runs on macOS (a Windows download is also available), integrates with your existing developer tools, and now ships with capabilities that push it well beyond the "autocomplete on steroids" category it once occupied.
Background Computer Use: Codex Works While You Work
The most consequential new feature is background computer use. Codex can now see your screen, move a cursor, click, and type — operating any app on your computer as an autonomous agent. Critically, multiple agents can run in parallel without interfering with whatever you're doing in other windows.
For developers, the immediate use cases are iterating on frontend changes, testing applications, and working in tools that don't expose an API. For anyone managing repetitive software workflows, it's a meaningful shift: the agent isn't waiting for you to paste something in — it's just doing it.
In-App Browser and Image Generation
The Codex app now includes a built-in browser. You can comment directly on pages to give the agent precise instructions, which is particularly useful for frontend and game development. OpenAI says they plan to expand browser command capabilities beyond localhost applications over time.
Codex also now generates and iterates on images using OpenAI's gpt-image-1.5 model. When combined with screenshots and code, it can produce visuals for product concepts, mockups, and UI designs within the same workflow — no switching between tools.
90+ New Plugins Across the Developer Toolchain
This update ships with more than 90 additional plugins, combining app integrations, skills, and MCP servers. New additions include Atlassian Rovo for Jira management, CircleCI, CodeRabbit, GitLab Issues, Microsoft Suite, Neon by Databricks, Remotion, Render, and Superpowers.
These aren't cosmetic integrations. The plugins give Codex contextual access to gather information and take action across the tools a development team actually uses — which is precisely what separates a useful agent from a toy.
Memory, Scheduling, and Long-Running Work
Two features here are worth understanding separately.
Memory allows Codex to retain useful context from previous sessions: personal preferences, corrections, and information that took time to gather. Future tasks are completed faster because the agent isn't starting from scratch every time.
Automations now support reusing existing conversation threads, which means context built up over time carries forward. Codex can also schedule future work for itself and wake up automatically — across days or weeks — to continue on a long-term task.
The practical example OpenAI gives is clarifying: Codex can identify open comments in Google Docs that need attention, pull relevant context from Slack, Notion, and your codebase, and serve you a prioritized action list at the start of your day. That's not a feature. That's a workflow replacement.
Full Software Development Lifecycle Support
The app now handles GitHub review comments directly, supports multiple terminal tabs, connects to remote devboxes over SSH (in alpha), and opens files in a sidebar with rich previews for PDFs, spreadsheets, slides, and docs. A summary pane tracks agent plans, sources, and artifacts across a session.
The through-line is continuity. Codex is designed so that writing code, reviewing changes, checking outputs, and coordinating with the agent all happen in one workspace, without context switching breaking your momentum.
What This Means Beyond Developer Teams
Here's where it gets interesting for marketing and growth leaders: Codex is increasingly less of a developer-only tool. The memory features, browser integration, plugin ecosystem, and scheduling capabilities describe something closer to a persistent digital operator than a code editor.
Teams already use Codex automations to land open pull requests, follow up on tasks, and monitor conversations across Slack, Gmail, and Notion. As the plugin library grows and computer use matures, the line between "AI coding assistant" and "AI that does knowledge work" continues to blur.
For growth teams thinking about how AI fits into their operational workflows — not just their content production — this category of tool is worth tracking closely. The question isn't whether AI agents will handle more operational tasks. It's which tools will do it reliably, and how fast your team can build the workflows to take advantage.
If you're thinking through where AI fits inside your marketing and growth operations, Winsome Marketing helps teams cut through the hype and build systems that actually work. Start the conversation here.