Serper

Serper occupies a deliberately narrow slot: real Google results, structured, fast, and priced so that an agent can query it hundreds of times per task without a budget review. Where Tavily fetches and cleans page contents, Serper gives you the SERP itself — titles, snippets, URLs, positions — and lets your pipeline decide what to fetch. Research agents commonly pair the two, or use Serper when Google's ranking signal matters (news recency, local results, shopping).

The economics are the story: legacy SERP APIs charged tens of dollars per thousand queries; Serper reset that to a fraction, which is why it spread through the agent framework ecosystem so quickly. It is plumbing rather than magic — no content extraction, no semantic search — but for “what does Google say, at scale, cheaply”, it is the industry's quiet default.