OpenAI Codex app update: workspace settings, Windows native support, MCP approval panels, and agent composer features
OpenAI shipped a Codex app update adding workspace settings, native Windows updater with MSIX and system tray support, and MCP custom approval panel handling. The automation composer gains app/file @mentions, subagent diff stats, and artifact cards for file citations. The release also includes a range of bug fixes covering multi-window issues, review panel behavior, and thread search performance.
Related guides (3)
Related events (8)
OpenAI Codex app update: realtime voice context and MCP panel stability fixes
OpenAI released a minor update to the Codex app adding a tooltip clarifying that realtime delegation messages use surrounding voice conversation as context. The update also fixes search navigation in long review files, prevents embedded MCP app panels from losing state during fullscreen changes, and resolves several desktop regressions including tray crashes and macOS hotkey issues.
OpenAI Codex app adds MCP shortcuts and review mention support
OpenAI's Codex app received a feature update adding MCP shortcuts in the composer—including install keyword suggestions and an MCP server submenu—alongside @mention and skill mention support in inline review comments. The release also includes improved rendering of MCP tool calls and Mermaid diagram error handling. The update signals continued integration of the Model Context Protocol into OpenAI's developer tooling.
OpenAI Codex app launches on Windows with parallel agent support
OpenAI has released the Codex desktop app for Windows, available to ChatGPT subscribers with Codex access. The app supports running multiple Codex agents in parallel with isolated worktrees and reviewable diffs that can be edited, discarded, or converted to pull requests. It integrates with the Codex CLI and IDE tooling, enabling a unified workflow across surfaces.
OpenAI expands Codex with in-app browser, computer use, PR workflow, and thread automations
OpenAI released a significant update to its Codex desktop app, adding an in-app browser for page-level feedback, computer use capabilities (macOS app control via vision and input), and deeper GitHub pull request integration. New workflow features include chat threads without project folders, scheduled thread automations for long-running tasks, and an artifact viewer for previewing generated PDFs, spreadsheets, and documents. Computer use is geo-restricted, excluding the EEA, UK, and Switzerland at launch.
OpenAI Codex app launches on Windows for Enterprise and Edu workspaces
OpenAI has released the Codex app for Windows, available to ChatGPT Enterprise and Edu workspace members. The app supports running multiple Codex agents in parallel with isolated worktrees and reviewable diffs, maintaining interoperability with Codex CLI and IDE surfaces. Permissions are unified under existing Codex Local and Codex Cloud policies, requiring no additional admin configuration.
OpenAI Codex gains Windows Computer Use and cross-platform remote control
OpenAI's Codex app now supports Computer Use on Windows, enabling the agent to operate Windows desktop applications by seeing, clicking, and typing. Remote control functionality has also been extended to Windows devices, allowing users to initiate and monitor Codex tasks from iOS, Android, or Mac. The update also adds thread coordination for local projects, expanded search across Codex threads, and improved Chrome context capture for Google Workspace apps.
OpenAI Codex app adds branch search, plan mode, and parallel approvals
OpenAI's Codex app received a minor update adding branch search in the branch picker, clearer guidance for entering plan mode, and support for parallel approvals. The release also includes unspecified performance improvements and bug fixes. These are incremental UX and workflow enhancements to OpenAI's coding agent product.
OpenAI Codex app adds MCP command palette and queued follow-ups
OpenAI's Codex app received a minor update adding MCP (Model Context Protocol) and personality actions to the command palette, along with a change to queue follow-up interactions by default. The update also includes unspecified performance improvements and bug fixes. The MCP integration is the most notable element, continuing the trend of MCP adoption across AI coding tools.