Tool surface
Core
web_search, web_fetch, web_extract
Use these first. Search ranks public hits. Fetch and extract read a page and return spans. Search does not fetch the hits, and it does not call web_research.
Advanced
web_cache, web_crawl, web_research, web_diff
Each one runs only when you call it.
| Tool | What you are asking for |
|---|---|
web_cache | Read or change the on-disk cache |
web_crawl | A same-origin crawl with depth and page caps |
web_research | Search, then fetch and extract the top pages (default 3, maximum 6) |
web_diff | Added and removed words between two texts or two URLs |
What is not a tool
Lookout has no profile argument and no hidden quality mode. It does not render a browser and it does not summarize with a model. A repeated search or fetch can come from the cache when useCache is left on. That cache is a replay of an earlier network response, not a substitute for the network.