Trawl – Scrape any site with natural language fields, not CSS selectors
Every scraper I've written has the same failure mode: it works for three months, a site redesigns, and my CSS selectors silently return empty strings. The data is still right there on the page — a human can find it instantly — but the scraper is blind. Trawl fixes this by splitting the problem. You describe what you want: trawl "https://books.toscrape.com" --fields "title, price, rating, in_stock" The LLM (Claude) looks at one sample item and derives a full extraction strategy — CSS selectors, attribute mappings, type coercion, fallback selectors. That strategy gets cached. Every…
What it does
In the maker’s words, at launch
Every scraper I've written has the same failure mode: it works for three months, a site redesigns, and my CSS selectors silently return empty strings. The data is still right there on the page — a human can find it instantly — but the scraper is blind. Trawl fixes this by splitting the problem. You describe what you want: trawl "https://books.toscrape.com" --fields "title, price, rating, in_stock" The LLM (Claude) looks at one sample item and derives a full extraction strategy — CSS selectors, attribute mappings, type coercion, fallback selectors. That strategy gets cached. Every subsequent page with the same structure is extracted with pure Go + goquery. No API calls, no token cost, full concurrency. The key insight: LLMs are good at understanding HTML structure, but you don't need them to extract 10,000 rows. Use AI for intelligence, Go for throughput. When a site redesigns, the structural fingerprint changes, the cache misses, and trawl re-derives automatically. You can preview exactly what it figured out: $ trawl "https://example.com/products" --fields "name, price" --plan Strategy for https://example.com/products Item selector: div.product-card Fields: name: h2.product-title -> text (string) price: span.price -> text -> parse_price (float) Confidence: 0.95 Some things that took real engineering effort: - JS-rendered SPAs: headless browser with DOM stability detection — polls until element count stabilizes and skeleton loaders resolve, scrolls to trigger lazy loading, auto-clicks "Show more" buttons - Multi-section pages: detects candidate data regions heuristically, target a specific section with --query "Market Share", scopes extraction via container selectors - Self-healing: monitors extraction health (% of fields populated), re-derives the strategy if it drops below 70% - Iframes: auto-detects and extracts from iframes when they contain richer data than the outer page Output is JSON, JSONL, CSV, or Parquet. Pipes cleanly: trawl "https://example.com/products" --fields "name, price" --format jsonl | jq 'select(.price > 50)' Written in Go. MIT licensed.
Does the same job
all alternatives →- LSLLM Scraper – turn any webpage into structured data2024 · github.com · ▲88
- RLRobust LLM extractor for websites in TypeScriptMar 2026 · github.com · ▲72
We've been building data pipelines that scrape websites and extract structured data for a while now. If you've done this, you know the drill: you write CSS selectors, the site changes its layout, everything breaks at 2am, and you spend your morning rewriting parsers. LLMs seemed like the obvious fix — just throw the HTML at GPT and ask for JSON. Except in practice, it's more painful than that: - Raw HTML is full of nav bars, footers, and tracking junk that eats your token budget. A typical product page is 80% noise. - LLMs return malformed JSON more often than you'd expect, especially with…
- IMI'm making an AI scraper called FetchFox2024 · fetchfoxai.com · ▲105
Hi! I'm Marcell, and I'm working on FetchFox (https://fetchfoxai.com). It's a Chrome extension that lets you use AI to scrape any website for any data. I'd love to get your feedback. Here's a quick demo showing how you can use it to scrape leads from an auto dealer directory. What's cool is that it scrapes non-uniform pages, which is quite hard to do with "traditional" scrapers: https://youtu.be/wPbyPSFsqzA A little background: I've written lots and lots of scrapers over the last 10+ years. They're fun to write when they work, but the internet has changed in ways…
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, March 2026
the whole month →
Switch from ChatGPT to Claude with import memory feature
AI · Mar 2026 · claude.com
