Browser Use

BU Ultrafast, Browser Use's fastest browser agent, is now on every Browser Use Cloud plan in the US, including the free plan. It was behind a waitlist until Oct 30, as announced on October 31.

The median Ultrafast task takes 22 seconds and costs $0.0064. The default agent on GPT-6 Luna takes 79 seconds and costs $0.0098, so Ultrafast is more than 3x faster and cheaper per task.

The figures are medians of completed runs from Oct 16 to 29, 2026, with at most 1,000 runs per project.…

Browser Use has published a tutorial on building a competitor price tracker in Python with Browser Use and SQLAlchemy, in a new guide. It scrapes three stores, keeps a price history in SQLite, flags changes, and runs on a schedule.

The example follows one product, LEGO set 77251, across LEGO, Target and Best Buy. It asks for each retailer's public USD item price. Each listing is compared only with its own previous observation, and changes are printed.…

Browser Use has published a guide to blocked headless browsers, comparing stealth plugins, managed browsers, residential proxies, verified bots, and official APIs.

Its useful distinction: changing an IP address doesn’t fix browser fingerprints or missing cookies, and stealth plugins don’t change TLS or IP signals. Managed browsers are paid, and Browser Use says access to every site isn’t guaranteed. It sells one of the browsers covered.…

Browser Use’s cloud automation guide recommends giving each queued job its own isolated browser, driven by Playwright or Puppeteer over CDP, the Chrome DevTools Protocol.

Cap workers at your plan’s concurrency limit, pace browser creation under its rate limit, and stop each browser when the job finishes. Closing the CDP connection alone doesn’t stop a Browser Use browser.

The guide also compares provider limits and costs.…

Carnegie Mellon University and Browser Use have released Odysseys2, a benchmark of 100 long tasks on the live web.

In their results, GPT-6 Astra at default effort fully completed 47 tasks. Opus 5.5 at high effort followed with 18. Full completion requires meeting every check for a task, rather than getting most of it right.

The team tested 15 agents with one attempt per task and 60 minutes per attempt, all using Browser Use’s BrowserCode harness and stealth browsers with CAPTCHA solving.…

Browser Use has a guide to fixing Playwright scripts that break when a site changes. The main advice: find elements by role, label, text or a test ID you control, and use Playwright's built-in waiting rather than fixed delays.

On sites you don't control, the guide covers manual repairs, Playwright's healer agent and Browser Use fallbacks. Healer patches need review. Browser Use fallbacks use model tokens, and the agent's actions should be checked.…

Browser Use published a guide to daily invoice downloads from a supplier portal with no API. It starts with asking the supplier for an export or feed, then checking the portal’s terms and your contract before automating access.

The guide compares Playwright or Selenium scripts, RPA and Browser Use Cloud. Browser Use recommends Playwright for one stable portal when a developer is available, and a browser agent for changing layouts or many portals, with paid tokens on every run and output that…

Browser Use published a guide to extracting structured JSON from websites with an LLM, comparing Firecrawl, Crawl4AI, Browser Use and Stagehand.

Its recommended starting point is one reusable JSON Schema. Fetch clean page text for static pages, and use a browser agent when a site needs logins, clicks or a JavaScript flow.

Validate every record in code, retry or flag failures, and keep the source URL with the result. Missing values should be null rather than guesses.…

Browser Use has a guide to connecting coding agents to your signed-in Chrome, covering Claude Code, Codex, Gemini CLI and Cursor. It also covers Browser Harness, which Browser Use makes.

One useful distinction: Playwright MCP’s extension mode gives an agent access to tabs you pick. Chrome DevTools MCP’s auto-connect mode and Browser Harness expose your whole running Chrome.

Playwright connections need approval unless you configure an extension token.…

Browser Use has published a guide to web agent benchmarks, comparing what they test and how they score it.

It separates controlled sandbox tests such as WebArena, live-site tests such as WebVoyager, and offline datasets such as Mind2Web that compare predicted steps with recorded human actions. The guide also points to benchmarks for screenshot-based agents, deep research and enterprise workflows.

Scores need context: Browser Use says changing only the judge model on the same 104 finished…