


Browser infrastructure that scales with you 🚀 Run scraping, automation & testing without managing browsers. browserless.io


Testing libraries like Playwright weren't built for automation. One interaction can mean thousands of messages to Chrome. That's why we built BAP, the Browser Automation Protocol. It's now in Python. We kept Playwright's method names and replaced what runs underneath them. Each call is one message, so scripts stay fast on a remote browser. The same script runs on our stealth browser when a site starts blocking you. pip install browserless-bap Python 3.11+, sync and async. Docs in the replies.

Smart Scrape just got its biggest update since launch. - Main content by default. No menus or footers. - JSON extraction from a schema or a prompt. - Click, type and scroll before capture - Batch Scrape for URL lists Plus caching, change tracking and more: na2.hubs.ly/H090XfY0

The Browserless MCP server can now pay at checkout with Stripe Link. You approve the exact amount, and the model never sees your card. /scrapers returns Zillow listings and property details as clean JSON in one API call. /agent/run can pause at a login or MFA prompt, let a person finish that step, then pick up where it left off. Full November changelog → na2.hubs.ly/H08ZXzp0

Enterprises are already handing real processes over to AI agents, and that still surprises me. Most of our new inbound at @browserless is exactly this. Not because the agents can't do the work. Because they're non-deterministic.


Testing libraries like Puppeteer are too chatty for automation. They can send hundreds, if not thousands, of messages to Chrome. BAP, the Browser Automation Protocol, sends far fewer. Joel explains why that makes it faster, and what owning the whole stack lets us build next.

Ever wondered where proxy vendors get their IPs? A lot of them come from apps on smart TVs, sometimes on devices whose owners don’t know. Joel’s advice is to ask any vendor two things before you buy. Where do the IPs come from, and does the device owner get paid?

Don't keep your secrets in environment variables. Railway's breach and the PHP exploit both landed this week. Joel's advice after Railway: move secrets out of env vars, and put a proxy between you and your LLM so you can see what's leaving.

Watch the full interview: youtu.be/eUMUpv_515A

Intelligent Audit tried 4 or 5 vendors before Browserless. "Everybody was either questionably secure, unsecure, or just completely out there." Browserless was the simplest to set up. They now run 1,000-1,500 UPS logins a day and file over 90% of claims through automation.

Connect your wallet here: browserless.io/account/integ…

Logged-in checkouts are the hard part of agent payments. Watch an agent on Browserless handle one end-to-end. With one prompt in Claude Code, the agent signs in with a saved profile, finds the top-up page, and waits for your approval in Stripe before anything is charged. You never typed a password or a card number. The saved profile handled the login, and Stripe issued a one-time card for that payment. US only, up to $50 per purchase.

Most agents reach checkout and stop. Nobody hands a model a real card number. Your Browserless agent can now pay with Stripe Link. You approve every charge, and each purchase uses a one-time card the agent never sees. Available for US users paying in USD, up to $50 per purchase. Docs: na2.hubs.ly/H08lMLW0

How a two-person team runs it all: browserless.io/customers/int…

"You guys actually spend time trying to help us debug our own code." Intelligent Audit logs into FedEx, UPS and DHL portals thousands of times a day on Browserless. When their own automation code broke, our engineers dug in and sent back snippets and working code.