OpenAI Operator.
Autonomous browser agent that drives real websites from a ChatGPT Pro pane.
What it is.
OpenAI's autonomous browser agent. Operator drives a real Chromium instance, takes screenshots, clicks, types, and submits forms. Shipped in January 2025 to ChatGPT Pro subscribers as a research preview. The Computer-Using Agent model handles vision and action in one stack — the first consumer-grade general computer-use product OpenAI has shipped.
Where it fits.
Workflows where a knowledge worker would otherwise repeat the same browser sequence — expense reports, supplier portals, travel booking, lead enrichment. Strong for individuals on the $200 Pro tier. Enterprise sits behind the upcoming Operator API and remains gated by procurement review.
- First consumer browser agent from a frontier lab
- Handles login walls and CAPTCHAs with human handoff
- Tight integration with ChatGPT memory and tools
- US-only and ChatGPT Pro gated until API ships
- Latency runs five to fifteen seconds per action
Frequently asked.
Is Operator available outside the United States?
Not yet. The research preview is US-only for ChatGPT Pro subscribers. International rollout follows the Operator API release.
Does Operator handle authentication and payments?
Yes for login flows with human handoff at the password prompt. Payment confirmation always returns control to the user before submission.
How does Operator compare to Anthropic's computer use?
Operator is consumer-packaged inside ChatGPT. Anthropic's computer use ships as a developer API under Claude. Same primitive, different distribution.
