Olostep Search endpoint

Turn a question into ranked web links.

The Search API (POST /v1/searches) returns deduplicated links with titles and descriptions — when you need discovery, not a full synthesized answer. For grounded Q&A and JSON schemas, use Answers; for content from a URL you already have, use Scrapes.

Try a natural-language query. The API returns deduplicated links with titles and descriptions — see Search API and Create search.

How it works

POST your natural-language query.

Read result.links (and optional hosted JSON).

Re-fetch by id or hand off URLs to scrape / batch.

Trusted by teams worldwide

Merchkit
Podqi
Khoj
Finny AI
Contents
Athena HQ
CivilGrid
GumLoop
Plots
Uman
Verisave
Relay
OpenMart
Profound
Centralize
Use Bear
Merchkit
Podqi
Khoj
Finny AI
Contents
Athena HQ
CivilGrid
GumLoop
Plots
Uman
Verisave
Relay
OpenMart
Profound
Centralize
Use Bear
Merchkit
Podqi
Khoj
Finny AI
Contents
Athena HQ
CivilGrid
GumLoop
Plots
Uman
Verisave
Relay
OpenMart
Profound
Centralize
Use Bear
Merchkit
Podqi
Khoj
Finny AI
Contents
Athena HQ
CivilGrid
GumLoop
Plots
Uman
Verisave
Relay
OpenMart
Profound
Centralize
Use Bear

From query to link list

Discovery API — pair with Scrapes or Batches when you need page bodies.

Step 1

Send your query

POST JSON with query — plain English, no URL list required.

Step 2

Receive links

Get a search object with result.links: url, title, description (deduplicated).

Step 3

Re-fetch by id

Store id from the response; call GET /v1/searches/{search_id} to load the same result later.

POST /v1/searches

{
  "query": "Best AEO startups 2026"
}

Built for discovery

Semantic search across the web, structured for agents and dashboards.

Semantic search

Expands your query into targeted retrieval across the web — not just keyword matching on one domain.

Structured links

Each item includes url, title, and description so UIs and agents can rank and filter quickly.

Hosted JSON

Full payloads may include json_hosted_url for downloading the complete result set when needed.

5 credits per search

Predictable cost per request — pair with Scrapes or Batches only for URLs you actually need.

What you can build

Use Search where a SERP-style link list beats scraping search HTML yourself.

Research & copilots

Surface candidate sources before deep reading or scraping.

Lead & market discovery

Turn a question into a shortlist of companies, articles, or product pages.

Pipeline front-door

Search → filter → enqueue URLs into batch jobs or /v1/scrapes.

Web search API

One query. Deduplicated links.

Optional json_hosted_url for the full payload. Retrieve again with GET /v1/searches/{search_id}.

RequestPOST /v1/searches
{
  "query": "Best Answer Engine Optimization startups"
}
Response200 OK
{
  "id": "search_9bi0sbj9xa",
  "object": "search",
  "created": 1760327323,
  "query": "Best Answer Engine Optimization startups",
  "result": {
    "links": [
      {
        "url": "https://www.adweek.com/media/…",
        "title": "Meet the 7 Hottest Startups…",
        "description": "1. Profound · 2. Bluefish · …"
      }
    ],
    "json_hosted_url": "https://olostep-storage.s3.us-east-1.amazonaws.com/search_9bi0sbj9xa.json"
  }
}

Frequently asked questions

Search vs Answers, Map, and how to chain with Scrapes and Batches.

How is the Search endpoint different from Answers?

Search returns a deduplicated list of relevant links (url, title, description) for your query — ideal when you want candidates to rank, filter, or scrape yourself. Answers runs a deeper agentic flow to synthesize a single response and optional structured JSON with source validation.

How is Search different from Map?

Map discovers URLs for one domain you provide (sitemaps + links on that site). Search runs a semantic query across the web and returns topically relevant pages from many domains.

When should I chain Search with Scrapes or Batches?

Use Search to propose URLs, then pass selected URLs to POST /v1/scrapes for full page content, or to a Batch job when you have a large list. Search gives discovery; Scrapes/Batches give extraction at scale.

What does a search response include?

A search object with id, query, and result.links. Each link has url, title, and description. The full payload may also expose json_content and json_hosted_url for the complete structured result per the API reference.

Can I retrieve a search again later?

Yes. After POST /v1/searches, keep the returned id and call GET /v1/searches/{search_id} to fetch the same search object again.

How much does Search cost?

Search costs 5 credits per request per Olostep pricing documentation. Check your dashboard for current balances and plans.

Try semantic web search

500 free credits to try — no credit card required.

Get search by id — API reference