Signals
OpenAI Agents API Now Drives a Hosted Browser for You

Agents built on the OpenAI Agents API can now operate websites without APIs inside an OpenAI-hosted browser. You no longer have to run your own browser fleet to let an agent click through a legacy portal; your app still owns approvals and sign-in.
What Changed
- Before: an agent that needed a site without an API meant running and patching your own headless browser stack.
- Now: the Agents API ships a computer use tool that runs the agent in a browser OpenAI hosts.
- Your application still decides which sites the agent may visit and handles any sign-in, so credentials and consent stay with you.
Who Should Care
- Builders whose agents stall at legacy portals, vendor dashboards, or internal tools with no API.
- Small teams that want browser-capable agents without owning browser infrastructure.
Who Should Not Care
- Teams whose compliance rules require self-hosted browsers or keep traffic off third-party infrastructure.
- Builders whose targets already expose solid APIs, where direct calls stay cheaper and more reliable.
The Verdict
The real shift is that browser access becomes a tool call instead of an infrastructure project. If an agent workflow is blocked on a site without an API, prototype it on the hosted browser this week and compare completion rate and cost with your current setup. If you need self-hosted browsers, keep your stack.
Take This To Your Agent
Copy a ready-to-paste investigation handoff that asks your coding agent to check this update against your product and recommend what to do next.