THE DUDE’S FIELD GUIDE
Playwright vs Puppeteer vs Selenium vs Browser Use
The quick answer
Choose Playwright for a repeatable browser-testing workflow, Puppeteer for focused Node.js browser scripts, or Selenium when WebDriver and existing infrastructure are important. Browser Use addresses a different need: a model deciding how to carry out a browser task. That flexibility introduces model cost and less predictable behavior.
This comparison summarizes documented capabilities. We have not run a comparative benchmark; fit depends on your workload and environment.
At-a-glance comparison
| Project | Best fit | Costs and services | Why consider it | Watch out for |
|---|---|---|---|---|
| microsoft/playwright | Repeatable tests across Chromium, Firefox and WebKit | Local tests need no paid model API | A test runner with assertions, isolation and traces | Selectors and expectations still need maintenance |
| puppeteer/puppeteer | Chrome/Firefox automation in JavaScript | Local scripts need no paid model API | A focused browser library for screenshots, PDFs and scripts | You assemble the surrounding workflow and checks |
| SeleniumHQ/selenium | WebDriver workflows and remote browser grids | Local use has no required service fee | Existing language bindings, tests and Grid infrastructure | Distributed execution adds operational work |
| browser-use/browser-use | Tasks where a model chooses browser actions | Default cloud/model route can be paid | Natural-language goals and adaptable navigation | Verify browser/model configuration and every consequential result |
Start with the outcome, not the agent label
A checkout regression test needs to fail reliably when the application is wrong. An exploratory assistant may need to work out an unfamiliar sequence of steps. Those are different requirements. Write down whether you need deterministic assertions, flexible navigation, browser coverage or simply a screenshot before choosing a tool.
What does local actually mean?
A browser can run on your machine while screenshots or page contents are sent to a remote language model. Check the browser location and model location independently. The Browser Use quickstart currently chooses cloud services by default; a local browser is an explicit configuration choice. Playwright, Puppeteer and Selenium can run ordinary scripts without involving a model at all.
How to evaluate a candidate
Pick a reversible task on a site you control. Run it with a fresh browser context and define an observable success condition. Include a slow page, a changed element and an expected failure. Inspect the trace or logs instead of treating a completed process as proof that the task succeeded. Compare maintenance effort as well as runtime.
When an agent is worth the extra moving parts
An agent can be useful when a task varies enough that maintaining a fixed script is costly. That is an editorial judgment, not a guarantee that an agent will handle every variation. Keep account permissions limited and put a review step before payments, deletions or public submissions. For a stable repetitive task, start by seeing whether an ordinary script is sufficient.
Common questions
Is Browser Use a free replacement for Playwright?
They solve overlapping but different problems. Browser Use adds model-driven decision making; its code license does not make model inference or its hosted browser free.
Do I need an LLM for browser automation?
No. Playwright, Puppeteer and Selenium can automate explicit browser steps without an LLM. An agent or MCP client introduces a separate model layer.
Sources and review
- microsoft/playwright documentation
- puppeteer/puppeteer documentation
- SeleniumHQ/selenium documentation
- browser-use/browser-use documentation
Check current installation instructions, licenses and service terms before choosing a tool. This guide’s review date changes only after an editorial recheck.
How we research and label evidence · Explore AI agents & orchestration