SerpHubAPI
For agent developers

AI agent search API

An agent that searches like a person — one query, ten links, read each page — wastes calls and still guesses. This API gives an agent what it actually needs: an answer it can cite, page text it can read, and several searches in one round trip.

What you get

Example request

curl https://api.serphubapi.com/v1/ai-search \
  -H "X-API-Key: shapi_..." \
  -H "Content-Type: application/json" \
  -d '{"q": "What changed in the latest EU AI Act guidance?", "provider": "auto"}'

Response points: the answer, the sources cited for it, and meta.credits_charged for the call. Billing follows the same rules as search: cache_first is half price, cache hits, failures and empty results are free.

Questions

What is the difference from plain web search?

Plain search returns ranked links. AI search returns an answer plus the sources behind it, so an agent can quote instead of guess. Page text (Extract) and batch search cover the other two things agents keep asking for: what a page says, and several searches at once.

What does it cost?

AI search per successful live call: DeepSeek 8 credits, Yuanbao 10, Tavily 48, Exa 21 (live status for every service is on the catalog page). Page text is 3 credits a page; batch queries are billed like single searches. Cache hits, failed calls and empty results are free.

How does my agent connect?

Two ways: the MCP server (list_services, search, fetch) for agent clients — see the install page — or the SKILL.md in the repository root for agents that call the API directly.

How do I start?

Sign up for 1,000 free credits, copy your key from the console, and call POST /v1/ai-search with provider auto, or POST /v1/search for ranked results.