Scout
Scout (@web) is the built-in agent for the web. It searches, reads pages, and answers with links to where it found things.
What’s the latest on the Artemis mission?
Find the opening hours for the Science Museum.
Who won the match last night?
It’s on from the start, with no keys or accounts needed. You can turn it off on its card on the Agents page if you’d rather your agents never searched the web. It can’t be renamed or removed.
What it can do
Section titled “What it can do”| It can | Notes |
|---|---|
| Search the web | Through DuckDuckGo, with no key. Add a BRAVE_API_KEY on Connections and it uses Brave Search instead, which is steadier and returns more. |
| Read a web page | Public sites only. Addresses on your own network are refused, at every redirect. It also reads JSON and PDF. |
| Read many pages at once | Up to 50, for research. It sees a short excerpt of each, about what it’s looking for. |
For research that takes many steps, it can work as a long-running agent while you carry on chatting.
Kept apart from your files
Section titled “Kept apart from your files”Scout never sees your files. A web page can try to talk an agent into leaking something, so the agent that reads the web has nothing to leak. Everything on a page is treated as data, never as instructions.
The System agent can also search and read the web, but only once you turn on Read web pages in its settings.