aptrouter

Verbatim, source-cited search — built token-lean for LLM agents. No LLM in the pipeline, so nothing is paraphrased and nothing is hallucinated.

Quickstart API reference

Why aptrouter

Verbatim citations

Every result carries a quoted source sentence and its identifier — grounded by construction, not LLM-rewritten. You can attribute every claim.

Token-lean by design

Abbreviated keys (d,t,y,a,s,q) and an X-Token-Budget header keep responses tiny — built for agents that pay per token downstream.

Prepaid, $0.005 per search

Pay only for what you call — each search debits one credit ($5 per 1,000). No plans, no seats, no monthly quota. Track spend with the /v1/usage endpoint; at a zero balance the API returns 402 until you top up in the dashboard.

One call

POST /v1/search with a bearer key returns ranked, cited results. See the API reference for the full contract.

First call

curl -X POST https://api.aptrouter.com/v1/search \
  -H "authorization: Bearer sk_live_…" \
  -H "content-type: application/json" \
  -d '{"query":"statins and breast cancer survival","limit":3}'

Each result is the highest-scoring source sentence plus its id, title, year, authors, and score. Opt into full abstracts with "fields":"abstract". Walk through it step by step in the quickstart, or read the full schema in the Search API reference.

Also from aptrouter

/search is GA — and it is the first of three products sharing one account and billing spine. /code, an autonomous cloud coding agent, and /design, an AI web-design agent with deterministic quality gates, are both in preview behind waitlists.