— AGENT-NATIVE SCRAPING

Give your agent the answer.

Turn a website's listings, products, or posts into a search API your agents can query in natural language. Scrape once, query forever. No embeddings pipeline, no vector database to babysit, no search index to wire up.

Start scraping Read the benchmark

— HOW IT WORKS

From URL to answer in three steps.

— 01 / DEFINE

Chat with the agent.

Tell it what to scrape. The agent picks the selectors, detects the fields, and validates the run.

Scrape acmehomes.co.uk for 3-bed houses in Bristol.

Found the listing page. Detected fields: title, price, bedrooms, features. Add anything else?

— 02 / INDEX

Trawley runs the pipeline.

Scrape, structure, embed, index. No infrastructure to provision.

  1. scrapehtml → structured
  2. embedstructured → vectors
  3. indexhybrid search ready
  4. re-scrapeon a schedule

— 03 / QUERY

Ask in English.

One endpoint. Structured, ranked results. Plug into Claude, Cursor, or your own agent.

bash
curl -G $TRAWLEY/v1/scrapers/$ID/search \
  -H "Authorization: Bearer $KEY" \
  --data-urlencode \
    "search=3-bed houses with a garden \
     under £500k in Bristol"

— WHAT'S IN THE BOX

Everything your agent needs from a scraper.

One platform for setup, scraping, indexing, search, and freshness. Each piece below is a capability you'd otherwise wire together yourself.

— 01 / SETUP

Agentic scraper setup

Chat to set it up. The agent picks fields, finds selectors, and validates the run.

— 02 / MAINTENANCE

Scraper doctor

Detects when sites change and autofixes selectors so your data keeps flowing.

— 03 / QUERY

Natural language search

Plain-English queries become filter, fuzzy, and vector search under the hood.

— 04 / QUERY

Filtered REST API

Typed filters and sorts for when you want explicit control. JSON in, JSON out.

— 05 / FRESHNESS

Scheduled scrapes and change tracking

Run on any cadence. Diff between runs to see exactly what changed.

— 06 / DEPTH

Nested page scraping

Follow links from a list page to grab structured details from each inner page.

— 07 / COMPATIBILITY

Built for real sites

JavaScript rendering. Pagination handled. Field types stay typed: numbers as numbers, dates as dates.

— 08 / ACCESS

Use your data anywhere

Chat with results in the Trawley UI. Export to CSV. Or hit the API from your stack.

Coming soon

— 09 / ROADMAP

Webhooks and MCP

Push results when scrapes complete. Create scrapers and query data from agent tools.

— THE QUERY LAYER

Natural language. All the precision of faceted search.

Most natural-language search bolts an embedding model onto a vector database and hopes ranking sorts it out. Trawley reads your query, picks the structured filters and ranges it can actually constrain on, runs semantic match for the rest, and ranks the result. You ask. Trawley figures out the most effective query.

— INPUT

3-bed houses with a garden under £500k in Bristol

— RANGE

bedrooms ≥ 3

Numeric bound. Indexed for exact retrieval.

— RANGE

price ≤ £500k

Numeric bound. Indexed for exact retrieval.

— SEMANTIC

features ~ "garden"

Vector match. Catches "south-facing garden", "patio garden", "garden shed".

— SEMANTIC

location ~ "Bristol"

Vector match. Catches "Clifton", "Bishopston", "Redland".

— OUTPUT

Ranked, structured results. The agent gets the answer, not the haystack.

— WHAT YOU DON'T HAVE TO BUILD

The pipeline you'd otherwise build.

— WHAT YOU'D BUILD

Scraping engine
Headless browser pool
Structured storage
Embeddings model
Vector database
Hybrid search index
NL query translator
Self-healing maintenance

Eight services. Eight integrations. Eight things to keep alive.

Trawley.

One API key. One thing to learn.

— SCRAPE ONCE, QUERY FOREVER

One scrape. N queries.

Re-extracting on every agent turn re-fetches the page, re-runs an LLM against the raw HTML, and re-pays the latency and cost every time. Trawley scrapes once, embeds once, indexes once, and serves every subsequent query from the index.

— PER-QUERY LATENCY

Trawley
—
Re-extract per call
—

Benchmark publishes 2026-06-08.

— COST · 1,000 QUERIES

Trawley
$—
Re-extract per call
$—

Benchmark publishes 2026-06-08.

— COST · 10,000 QUERIES

Trawley
$—
Re-extract per call
$—

Benchmark publishes 2026-06-08.

— PRICING

Self-serve. Set pricing. No demo call.

— FREE

$0 /mo

For exploring Trawley.

  • 100 page credits/mo
  • 1 scraper
  • 50 search API calls/mo
  • 30-day retention
Start free

— STARTER

$39 /mo

For small projects.

  • 2,000 page credits/mo
  • 5 scrapers
  • 2,000 search API calls/mo
  • Scheduling
  • 90-day retention
  • 2 team members
Start Starter
Most popular

— PRO

$99 /mo

For growing teams.

  • 10,000 page credits/mo
  • 15 scrapers
  • 10,000 search API calls/mo
  • Scheduling
  • 1-year retention
  • 5 team members
Start Pro

— BUSINESS

$249 /mo

For large-scale operations.

  • 50,000 page credits/mo
  • 50 scrapers
  • 50,000 search API calls/mo
  • Scheduling
  • 2-year retention
  • 15 team members
Start Business

Every paid plan is set, self-serve, and cancellable in one click. No demo call required.

Build your first scraper in an afternoon.

Free to start. No card required. Cancel any paid plan in one click.

Start scraping