OpenAI Ships Codex Mobile — Steer Long-Running Agents from Your Phone
Codex, OpenAI's agentic coding environment, is now in your pocket. Rolling out in preview on iOS and Android across all plans — including free tiers in all supported regions — the feature lets developers monitor, approve, redirect, and launch tasks on a desktop Codex session directly from their phone. Files, credentials, and permissions stay on the local machine; the handset becomes a remote control for a long-running code agent. Three independent sources covered the launch the same day.
What the Source Actually Says
The mechanics are straightforward: Codex keeps running on the desktop while the developer steps away. When it finishes a task or needs a decision, it fires a phone notification. From there you can prompt further, approve a pending command, or redirect entirely — without sitting back down. The primary source (AI Search weekly roundup) spells out the use case directly: "Sometimes [Codex] takes like 5 to 10 minutes on a certain task, especially if it requires multiple steps or some really long reasoning. Instead of waiting for it, you can just go off and do your own thing and then once it's done, it'll send a notification."
The current constraint is macOS-only on the host side; Windows support is described as "coming soon." The NLP Newsletter's AI Agents Weekly simultaneously reported that Codex also gained a hooks system for lifecycle customization — directly mirroring Claude Code's hook architecture — suggesting OpenAI and Anthropic are converging on the same extensibility standard for coding agents.
Practitioner reaction on X confirms the release landed with impact. OpenAI president @gdb called the Codex app "in a category of its own." Developer @swyx coined the phrase "agentic excel on mac" during a live demo. @mattshumer_ immediately repurposed a dedicated Mac Mini as an always-on Codex devbox, and used Codex itself to one-shot a mobile companion app for his own product with "almost no effort."
Strategic Take
The async remote-control model changes the calculus for agentic coding work — long-horizon tasks no longer chain the developer to a desk. For teams evaluating coding-agent tooling, the hooks launch is the quieter signal: convergence between Codex and Claude Code on extensibility architecture means integrations built for one are increasingly portable to the other.