Each agent has its own web browser. It runs in the cloud and is separate from your own browser.
Watch and steer
- The Browser button at the top of an agent’s screen shows whether the browser is ready, live or working.
- The browser panel opens by itself when a task starts browsing. You can also open it with + then Show the browser.
- In the panel you can type an address or a search into Search or enter a web address, go back, forward or reload, open tabs, and click, type or scroll on the page yourself.
- Downloaded files lists what the browser has downloaded.
Signing in to sites
Password fields on a page are marked private. Use Enter privately to fill one: the value goes only to the page and is not shown or saved in the chat.
The agent’s browser keeps its own cookies and sign-ins for that agent. It never uses your own browser’s profile.
Limits
- It opens public https:// pages only.
- It can have up to five tabs open.
- Files it uploads to a page or downloads from one can be up to 500 KB.
Sites that block automated reading
This site refuses automated reading. Open it in the browser instead.
- Ask the agent to open the page in its browser.
This site blocks automated readers, including the browser reader. Open it yourself and paste the part you need.
- Open the page in your own browser and paste the part you need into the message.
