"If Browserless were not there, they'd have to move the case file by hand, which would be crazy."
One court case file can hold 400 to 450 entries. Every entry means a PDF downloaded, uploaded and logged. Pragma Legal does that in the background instead, on a session that stays logged in long enough to finish the file and reconnects if it drops.
Full story in the comment.
Getting past bot detection once is easy. Keeping it that way isn't.
BAP is a stealth-first SDK. If you've written Puppeteer or Playwright, you already know the methods. The browser runs on our side, so there's no stealth patch to keep current.
Residential proxies are built in, routed by country. Typing runs at human pace. When a challenge still shows up, one solve() call clears it.
TypeScript is out now, shaped like Puppeteer. Python is next, shaped like Playwright.
Free tier, no card.
npm install @browserless.io/bap-ts
"I am not a programmer. I am actually a lawyer."
HernΓ‘n Napolitano set up Browserless himself in about an hour and a half, with an AI assistant pointed at our API docs.
It signs into the court portal as the lawyer and pulls their case files out. That is the step his whole app runs on.
BAP, our new SDK, is out in TypeScript. Puppeteer-shaped API, with managed stealth, proxies and CAPTCHA solving underneath. Python is coming soon.
Request logs are live in your dashboard. Every run shows where it ran, how long it took, what it cost, and the error that killed it.
Live URLs now work on every plan, including Free.
Read our changelog β https://t.co/uI63OW2WPf
Now you can see exactly why a request failed.
Request logs are live on Browserless. Every request you send now shows up with the whole story. Which endpoint, how long it ran, the error that killed it, and the units it burned anyway.
Here's the tour.
Claude can write your job application answers. It can't browse LinkedIn job boards or fill out the form.
Give it a real browser with our MCP server, and it does the rest: searches LinkedIn for new postings, opens the application, reads your resume, fills it, submits.
Watch it work.
Introducing the Agent Skill Catalog. 378 skills for 344 sites.
Agents are smart but completely amnesiac. So we gave them a memory.
When our agent works out a hard site, we write down what it learned. The URL that skips the UI, where the JSON lives, how to get past the bot wall. Then your agent starts where ours finished, so the run is faster.
Try it with the Browserless MCP.
Two ways to make running browsers cheaper.
Rewrite the browser and give things up. Or run the real one on cheaper silicon.
@Cloudflare just picked the first one. @browserless is testing the second.
You can now create multiple API keys on a single Browserless account.
Name them per environment or per app manage them all in your dashboard. They share your account's units and concurrency.
Revoke one and it stops instantly. The rest keep working. Rotate keys with no downtime.
The hidden cost of running your own browsers isn't just the ~$20 pods. It's the balancers, the scaling, and the hours of engineering time that never quite make it reliable.
Norbert from Dune breaks down the real math behind DIY browser infrastructure.
Read the full story here: https://t.co/VLvkPDW3W6
New /agent/run endpoint spins up an AI browser agent from one call. Send a task in plain English, get clean JSON back.
Multiple API keys per account are available, so every app or teammate gets its own. You can also scrape mobile sites as a real Android device.
Full August changelog β https://t.co/2aM31JgQvq
Reliability isn't a nice-to-have when your product runs on data.
Norbert Durcansky at @Dune went from ~10% of browser requests failing or retrying to zero after switching to @Browserless. That's the difference infrastructure makes.