OpenAI's Codex Desktop adds computer control and browser automation beyond coding
OpenAI's Codex Desktop can now control your computer, run background automations, and includes an in-app browser with click-to-select elements. The update adds automation memory across sessions and access to over 100 curated plugins, though the computer control feature is MacOS-only and unavailable in the EU.
OpenAI's Codex Desktop adds computer control and browser automation beyond coding
OpenAI announced a significant expansion of Codex Desktop, adding computer control capabilities, an in-app browser, and persistent automation memory. The update positions the tool beyond pure code generation, though OpenAI confirms it remains targeted at developers.
Computer control limited to MacOS
Codex Desktop can now operate applications in the background while users work in other programs. The computer control feature is MacOS-only at launch and not available in the European Union. No Windows or Linux support timeline was provided.
Browser automation with click-to-select
The update includes an in-app browser with the ability to click elements directly rather than describe their location. Users can select a specific page element and issue commands like "change this to that" without explaining positioning.
Automation memory and task continuity
Codex Desktop now maintains memory across sessions, addressing what OpenAI acknowledged was a "particularly problematic" limitation requiring the app to be brought up to speed on every relaunch. The system can "wake up automatically to continue on a long-term task, potentially across days or weeks," according to OpenAI.
Automations can be added to existing conversational threads, preserving context from earlier interactions. The app proactively suggests continuing previous workflows on launch.
Image generation and developer features
Codex Desktop adds ChatGPT's image generation capabilities for automated chart and diagram creation within workflows. Developer-focused updates include:
- GitHub review comment support
- Multiple terminal tabs (no color-coding yet)
- SSH connections to remote devboxes (early testing)
- File previews for PDFs, spreadsheets, slides, and documents
- Summary pane tracking agent plans and sources
100+ curated plugins
OpenAI ships Codex Desktop with access to over 100 plugins combining skills, app integrations, and MCP servers. The company curates plugins before availability, differentiating from Anthropic's Claude approach which faced malware issues from user-contributed skills.
Availability and usage
Codex Desktop is available to any OpenAI tier with Codex access on Mac and Windows. Computer control remains MacOS-exclusive. OpenAI warns that running extended automations will consume token allocations faster.
According to OpenAI, 80% of its staff use Codex, including non-programmers, though the company maintains the primary target remains developers.
What this means
Codex Desktop's expansion into computer control puts OpenAI in direct competition with Anthropic's Claude and other AI automation tools. The MacOS-only limitation for the headline feature suggests technical or safety constraints OpenAI hasn't resolved for other platforms. The EU restriction likely reflects regulatory concerns around autonomous computer control. The real test will be reliability—click-to-select and multi-day automations are impressive in demos but need to work consistently in production environments.
Related Articles
OpenAI upgrades ChatGPT Chrome extension with multi-tab awareness and ChatGPT Work site control
OpenAI has upgraded its ChatGPT Chrome extension with a sidebar that reads across open tabs. Through ChatGPT Work, it can also control Chrome tabs to send messages, fill out forms and download files. TechRadar tested the form-filling feature and found it useful but in need of supervision.
OpenAI opens Decisions API to all customers, cutting latency to 100-180 ms on a reconfigured ChatGPT Luna
OpenAI made its Decisions API available to all customers on Thursday, running a reconfigured ChatGPT Luna that returns choices with confidence ratings in 100 to 180 milliseconds, according to OpenAI. The service competes with TypeSafe's Jev, which ranks higher on two public decision-model indices.
OpenAI agents on GPT-Rosalind analyze mouse brain CRISPR atlas in 32 minutes, versus months by hand
OpenAI, PerturbAI and Scripps Research deployed thousands of GPT-Rosalind agents to analyze a whole-brain CRISPR atlas covering millions of mouse cells. The full analysis took 32 minutes; the team estimates manual work would take several months to a year.
Google Cloud adds three OpenAI GPT-6 models in private preview, alongside Gemini and Claude
Google Cloud has begun offering three OpenAI models, GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna, in a private preview for select enterprise customers. Google says this makes it the only major cloud provider hosting Gemini, Claude and OpenAI models on one platform. Pricing, context windows and benchmark scores have not been disclosed.
Comments
Loading...