Try Trawl Before You Sign Up

Try It Before You Sign Up

Every scraping tool asks you to create an account before it shows you anything. We did too, and it was the wrong order. You cannot judge whether a scraper survives a real page from a screenshot on a landing page.

So we put the real wizard on a public URL. No account, no credit card, no email. Go to trawl.me/try and watch a scraper get built.

That is the headline of this round. Behind it sit five weeks and four release cycles shipped since v2.0, now live as API v2.3.0, web app v2.9.0 and CLI 3.8.0. Two themes run through all of it: remove the price of a first look, and make the free plan good enough to build on.

The Real Wizard, on a Public Route

The /try page is not a video and not a mock. It is the same wizard a signed-in user gets, running as a guest. Here is what happens in the ninety seconds it takes.

You pick one of three ready-made targets. The wizard opens the page and shows you what it sees, screenshot included, so you are not trusting a progress bar. It reads the layout and proposes the fields it thinks you want, in plain language, with what each one caught. You get back a table of clean rows, the same table a scheduled run would put in your inbox every morning.

• 🌐 Web app · Public demo: no account, no card, no email gate. The three targets cover what people ask about first, and they run from prepared data, so the demo is instant and cannot fail on you while you are still deciding whether to care.
• 🌐 Web app · A gentler first visit: the home page takes a target directly, the run reads as a conversation instead of a wall of logs, and if you like what you see the same session carries into signup. Nothing is retyped.

The honest trade: because it runs on prepared data, /try shows you the shape of the product, not a live scrape of your own URL. That one takes an account, because a live scrape spends real browser time on a real defended site. If you want to know what that costs before spending it, we explain why we price on compute rather than seats.

Curious how a scrape gets past a wall in the first place? We wrote up the proxy tier ladder.

API Keys Are No Longer a Paid Feature

Putting API keys behind a paid plan made a certain kind of sense and cost us the users we most wanted. A developer trying a scraping tool does not click through a UI. They script it, once, badly, to see if it works. So the whole loop now runs on the free plan:

$ export TRAWL_API_KEY=trawl_xxxxxxxx
$ trawl create https://example.com/rooms --prompt "nightly rate, room type, availability"
$ trawl data 66f1c4e2a9b3 --json

{
  "type": "success",
  "message": "scrap data",
  "data": [
    { "roomType": "Double", "nightlyRate": 129, "availability": "available" },
    { "roomType": "Suite",  "nightlyRate": 249, "availability": "sold out" }
  ]
}

The structure around the data is ours. The fields inside it are yours, defined by the sentence you typed. That last command reads the most recent run and costs you nothing, so it is safe to call from a scheduled job or a dashboard.

• 🔗 API · Keys on every plan: create a key and check whether the data fits your pipeline before any card is involved.
• ⌨️ CLI · Keys work in the terminal too: an agent working in your terminal no longer borrows your login. A key only reaches the scrapers you granted it, and revoking it does not sign you out everywhere.
• ⌨️ CLI · Readable by machines: one command prints everything the tool can do, as data. An agent learns the whole surface in one call instead of reading a manual and guessing.
• 🔌 MCP · Better visibility: we can now see which tools agents actually reach for, instead of guessing which ones matter.

Want the same thing running every morning? Start free, no credit card.

Credits and Invites for Everyone

A free plan that runs out before you learn anything is a demo with extra steps. Two changes fix that.

• 💳 Billing · 3,000 launch credits: the free grant is now large enough to build something real, absorb a few blocked runs, and let self-healing prove itself on a page that changes.
• 🌐 Web app · Invites for every user: invites used to be an internal tool. Now every account has them, with a nudge after your first successful run and a line in the digest email. Both sides get credits, and you get told when yours lands.
• ⌨️ CLI · A quiet reminder: a one-line invite tip after a run, rate-limited so it stays out of your way.

The Wizard Talks While It Works

A first scrape on a defended site takes real time. The old spinner hid that and it cost us trust, so the wizard now narrates.

• 🌐 Web app · Live progress: each stage appears as it happens instead of resolving into one silent wait. You see the page load, the fields proposed, the test run.
• 🌐 Web app · A conversation you can scroll: the chat stacks instead of replacing itself, screenshots and results inline, so you can go back through what it decided and why.
• 🌐 Web app · Sensible defaults: a scraper built here arrives with a weekly schedule and its email digest on. Building one and never scheduling it was the most common way to end up with nothing.

Quieter Under the Wall

The unglamorous half. Most of this you will never see, which is the point.

• ⚙️ Platform · Fewer blocks, less waiting: the browser we drive now looks ordinary in the small details anti-bot systems check, which is where most scrapers give themselves away. We also remember what worked on a site last time instead of rediscovering it on every run.
• ⚙️ Platform · The hold-to-continue walls: some sites now guard themselves with a button you have to press and hold. We recognise those and attempt them. We are not claiming we beat them: that kind of wall changes its defence between visits, so the honest status is "we try", and we are still measuring how often it works.
• 💳 Billing · Honest cost: what a run costs now includes the infrastructure behind it, not only the visible part, and a toggle lets you read it either way. Alongside a broad security pass. Nothing exotic, all of it overdue.

FAQ

Do I need an account to try Trawl?

No. Go to trawl.me/try and run the demo wizard as a guest. It builds a real scraper against one of three ready-made targets. An account is only needed to point it at your own URL, because that spends real browser time on a real site.

Can I use the API on the free plan?

Yes. As of API v2.3.0, API keys work on every plan, including free. Create one in the app and use it from your own code or from the CLI, with no card on file.

How do I point an AI agent at the CLI?

Set the environment variable TRAWL_API_KEY to your key and the CLI works as that key rather than as you. Then run trawl spec --json and the CLI describes everything it can do as data, so the agent does not have to read documentation. A few account-level commands still require a normal login, on purpose.

How many free credits do I get?

3,000 on signup. Invites add more on both sides, and every account can send them. Credits track browser time, proxy quality and AI calls, which is why we price on compute rather than seats.

Does it get through the hardest anti-bot walls?

Sometimes, and we would rather say that than round up. The toughest sites now use press-and-hold challenges that change between visits. We recognise them and try, we have not confirmed a stable success rate, and we are still measuring. If you want the full picture of why scrapers fail, we wrote up what actually breaks one in production.

That is the round. The shortest path in is trawl.me/try, which costs you nothing and about ninety seconds. If it holds up, the free plan now has enough credits and enough room to answer the only question that matters: does this survive on your page, tomorrow morning, without you.

Written by Pierre | August 2026