toolcall.
ToolsAug 30, 2026, 08:25 UTC

OpenAI gives ChatGPT Work a cloud browser for signed-in tasks

The agent can use a remote browser, pause for secure sign-in, and ask before consequential actions such as payments or bookings.

OpenAI has expanded ChatGPT Work with a cloud browser that can carry out tasks on public and signed-in websites while keeping account credentials out of the model’s view.

The help page says the browser runs on a separate cloud computer. When a site requires login, ChatGPT pauses and asks the user to enter credentials through a secure sign-in form. OpenAI says the username and password go directly to the remote browser, are not visible to the model, and are not stored by ChatGPT. A separate review model checks sign-in requests for signs of phishing or deception.

The feature is available in ChatGPT Work on paid plans in supported regions, with rollout and workspace-policy limits. OpenAI says it can help with tasks such as checking availability, tracking packages, comparing local stock, finding appointments, working through signed-in utility accounts, or updating records in business tools.

The important boundary is confirmation. ChatGPT is designed to ask before actions that create financial, legal, account, or other hard-to-reverse commitments, such as confirming a booking or making a payment. Some websites may still block automated browser agents, and OpenAI tells users to review sensitive steps and results before relying on them.

Sources

Mentioned

ai-agentsautomationchatgptprivacy