Use cases
What people build with it
Two endpoints, and most of what gets built is a combination of them. Here are thirty of them.
30 of 30 use cases shown
- Search + Extract
Ground an LLM in the live web
A model's training data is frozen and its guesses are confident. Search returns the pages that actually answer the question, already stripped to readable prose, so the context you paste into the prompt is evidence rather than recall.
- Search + Extract
Give an agent a web tool
Agents do not scan a page of blue links. They need the text, and they need to decide whether to trust it. Every result carries its ranking broken into six signals, so an agent can weigh a source instead of taking the top hit on faith.
- Search + Extract
Watch what changes about a topic
Run the same query on a schedule and diff the results. New domains appearing, a page's body text changing, a claim being quietly edited: all of it shows up when you hold the full text rather than a snippet.
- Search + Extract
Enrich a record from a company name
Turn a row in a CRM into something a human would have spent ten minutes finding. Search the name, read the site, and pull the description, the market, and the language they use about themselves.
- Extract
Build a clean training corpus
Raw HTML is mostly navigation, cookie banners and footers. Feeding that to a model teaches it about cookie banners. Extraction returns the article and nothing else, as text or Markdown, with a word count so you can filter thin pages before they reach your pipeline.
- Search + Extract
Check a claim and cite the source
A sentence is easy to generate and hard to verify. Search the claim, read the pages that come back, and return the passage that supports or contradicts it, with the URL it came from.
- Search
Follow a story as it develops
Filter by recency and a query stops returning the definitive article from 2019. Useful when what you need is what happened this week, not what is most linked overall.
- Extract
Turn any page into Markdown
Documentation, a blog post, a changelog, a spec. Paste the URL and get Markdown with the headings, lists and links intact, ready to commit, index or paste into a prompt.
- Search + Extract
Research before a decision
Deep tiers widen recall so the answer is not decided by whichever source happens to rank first. Useful when being wrong is expensive and you need the coverage to be defensible.
- Answer
Answer questions from current docs
Support answers go stale the moment a product ships. Reading the live documentation at answer time means the reply reflects the version the customer is actually using.
- Search + Extract
Qualify a lead before the call
Ten minutes of tab-opening before every call does not scale past a handful a day. Search the company, read the pages that describe it, and arrive with their positioning, their market and their own vocabulary already in front of you.
- Search + Extract
Size a market from public sources
Market figures are scattered across reports, filings, press releases and trade press. Deeper tiers widen recall so an estimate rests on several independent sources rather than the first analyst who ranked well.
- Extract
Track how competitors price
Pricing pages change quietly. Reading the page on a schedule and holding the full text means a tier being renamed, a limit being cut, or a plan disappearing all show up as a diff rather than a surprise.
- Search + Extract
Read hiring as a growth signal
Job listings are the earliest public signal of a strategy change. A company staffing a new team is telling you about a product six months before the launch post.
- Search + Extract
Build a content brief that cites sources
A brief written from ten titles is guesswork. Read the pages that actually rank, and the brief carries the arguments they make, the questions they answer, and the gaps none of them fill.
- Search + Extract
Find where you are being talked about
A snippet tells you a page mentions you. It does not tell you whether the mention is a recommendation, a complaint or a passing aside. Reading the page does.
- Extract
Summarise a page without guessing
Summary quality is mostly input quality. A model handed raw HTML spends its attention on menus and disclaimers. Handed the article, it summarises the article.
- Extract
Keep an internal knowledge base current
Internal wikis decay because nobody re-reads the sources. Extract the pages a document depends on, hold the text, and you can tell when the source changed under you.
- Search + Extract
Check a vendor before you sign
The pages that matter in procurement are the ones nobody links to. Search for them, read them in full, and pull out the clauses that decide whether this is signable.
- Search + Extract
Gather sources for a review
A literature search that leans on one index inherits its blind spots. Fanning the query across many independent sources and merging on a canonical URL produces a list that is broader and less duplicated.
- Extract
Compare products on their own claims
Comparison content is usually written by whoever is selling. Reading each vendor's own pages gives you their claims in their words, which is a fairer basis for a comparison than a third party's table.
- Search
Track a portfolio company
News about a company you hold arrives unevenly and rarely on schedule. A recurring query weighted toward recency turns that into something closer to a feed.
- Extract
Watch a dependency's release notes
Breaking changes are announced on pages nobody re-reads. Extracting the changelog on a schedule and holding the Markdown means a new entry is a diff rather than an outage.
- Search + Extract
Read the room on a topic
Sentiment drawn from headlines is sentiment about headlines. Reading the pages gives a model the argument being made, which is the thing you wanted to measure.
- Extract
Keep a copy of what a page said
A URL is not a record. If a claim matters, keep the text as it was when you relied on it, in a format that is still readable years later.
- Search + Extract
Write a brief for someone joining
Most of what a new joiner needs about a market is public and scattered. Search the space, read the landmark pages, and hand them a brief with the sources attached.
- Search + Extract
Curate a newsletter
Curation is mostly reading. Pull the week's results for the topics you cover, extract the bodies, and write summaries from the article rather than from its headline.
- Search
See what ranks for a query, and why
Most search APIs hand back an order and no explanation. Every result here carries its score broken into consensus, position, term matching, proximity, host authority and freshness.
- Extract
Index another API's documentation
Integrating against an API usually starts with reading its docs. Extracting them to Markdown gives you something you can index, search locally, or hand to a model as context.
- Search + Extract
Research beyond one language
Querying one language returns one region's view. Fanning across independent sources reaches regional pages, and every extracted page reports the language it was written in.
Start with $10 of credit
No card required. The free grant renews every month, whether or not you ever top up.