Summary of an article published on The Web Scraping Club on 2026-04-30 (for paid subscribers). Read the full article
Every web infrastructure company seems to be shipping a browser this year, one built to be driven by an AI agent and to look human while doing it. We wanted to know if any of them works against a serious anti-bot. Agents need current data, and most of it is reachable only through a browser. No tool bypasses every anti-bot on every site. What matters is what works on a given target, today, at what cost.
We picked leroymerlin.fr, which runs DataDome standalone and exposes its risk score in response headers. The task has four steps: open the homepage, search for a product, open the first result, add it to the cart. We tested four open-source projects and seven commercial products, ten runs each, using French residential proxies from Bright Data and Geonode where the tool allowed it. We also cover why proxy vendors now sell managed browsers.
What’s inside
- Why most code sold as agentic is workflow code with an LLM in a slot, and what separates it from a real agent.
- Why leroymerlin.fr is a clean target: DataDome alone, verbose headers, and no login step, so OTP friction does not confound the test.
- The open-source group: Camoufox, Pydoll, Scrapling in two modes, and RayoBrowse, with the proxy authentication problem we hit on one of them.
- The commercial group: Browser Use in CDP and agent modes, Browserbase, Browserless, ZenRows and Bright Data Browser API, with free tier limits that shape what a free user gets.
In the full article: Results per tool and per run, the DataDome risk scores, the fixes we made before the numbers made sense, and the code in the paid GitHub repository.