CL-SN55 941 —
GPT-61S 890 —
GPT-56T 861 —
GPT-6S 837 -1.2%
CL-OP55X 831 -1.4%
MUSE-SPK 804 —
QWEN-38X 795 —
GPT-6A 782 —
CL-OP5H 724 -1.6%
CL-OP5X 722 -1.6%
KIMI-K3X 718 -0.1%
CL-OP46H 707 -1.7%
CL-FAB5H 705 -7.4%
CL-OP55H 705 -1.7%
CL-OP47H 696 -1.7%
GEM-37FH 661 -9.7%
GEM-38FH 656 -4.8%
GPT-56S 578 -0.5%
CL-OP47 543 -0.5%
INKL 531 —
GEM-3P 494 -0.2%
CL-OP46 490 -0.2%
GEM-31P 485 -0.4%
CL-OP48 483 -0.2%
GPT-6L 436 —
CL-SN55 941 —
GPT-61S 890 —
GPT-56T 861 —
GPT-6S 837 -1.2%
CL-OP55X 831 -1.4%
MUSE-SPK 804 —
QWEN-38X 795 —
GPT-6A 782 —
CL-OP5H 724 -1.6%
CL-OP5X 722 -1.6%
KIMI-K3X 718 -0.1%
CL-OP46H 707 -1.7%
CL-FAB5H 705 -7.4%
CL-OP55H 705 -1.7%
CL-OP47H 696 -1.7%
GEM-37FH 661 -9.7%
GEM-38FH 656 -4.8%
GPT-56S 578 -0.5%
CL-OP47 543 -0.5%
INKL 531 —
GEM-3P 494 -0.2%
CL-OP46 490 -0.2%
GEM-31P 485 -0.4%
CL-OP48 483 -0.2%
GPT-6L 436 —
← Back to feed

OpenAI Brings Codex to Mobile: 4 Million Weekly Users Can Now Direct AI Coding Agents From Their Phones

OpenAI shipped Codex in the ChatGPT mobile app on May 14, framing it as the missing link for a new category of developer workflow: async collaboration with a coding agent running on a laptop, Mac mini, or managed cloud environment while the developer is away from their desk.

More than 4 million people use Codex every week. The mobile tab connects to any machine where Codex is running and loads live state: active threads, terminal output, diffs, test results, and pending approvals. A secure relay layer routes updates back to the phone without exposing the host machine directly to the public internet.

What it does

The mobile interface is designed for thread management rather than initiating new sessions from scratch, though that is also supported. From a phone, a developer can:

  • Review what Codex found or produced while they were offline
  • Approve or reject pending commands before they run
  • Change direction on an active thread without opening a laptop
  • Switch between models mid-session
  • Start new work if something comes to mind

Files, credentials, permissions, and local setup stay on the machine. Only output — screenshots, terminal logs, diffs, test results — flows back to the phone in real time.

Why it matters

The mechanics reflect a structural shift in how coding agents are being used. Codex tasks increasingly run for extended periods: investigating a bug, refactoring a module, generating a test suite. Those runs accumulate approval bottlenecks and decision points that previously required a developer to return to a workstation.

OpenAI is positioning Codex as infrastructure for continuous work rather than a session-bound tool. Keeping a developer reachable during a long-running task changes the effective throughput of a single agent thread.

The timing is competitive. Anthropic’s Claude Code has a similar remote-access capability via its web interface, and the developer tooling market is increasingly defined by which platform makes it easiest to stay connected to running agents across devices.

Availability

Codex in mobile is in preview. The secure relay syncs any signed-in ChatGPT account that has Codex enabled. OpenAI has not announced GA timing.

At 4 million weekly users, Codex is OpenAI’s fastest-growing developer product by active user count. The company is racing to extend that lead with Anthropic’s Claude Code also adding enterprise features at pace.