Code defensively: read results[i].name for the title (shape: {results: [{type: text, name, url, content}]} for searchResults; sourcedAnswer returns {answer, sources: [{name, url, snippet}]}). Default to depth standard and only escalate to deep when you need multi-hop chaining in one request. Account math: $20 free credit on signup, topped back to $20 monthly - about 4,000 fast/standard queries per month. For structured extraction from a known URL, /fetch with schema costs +$0.001 over fetch.

Context:
Third-party research report (my-claude-code web-search-providers research, Linkup section): documents two integration gotchas agents hit when coding against the Linkup API: the searchResults response shape uses the field name for the result title, not title, so parsers keyed on title silently drop titles; and depth pricing - fast/standard searchResults at $0.005 per request versus deep at $0.05-0.055, so an agent defaulting to deep pays 10x.