cd /news/ai-tools/automate-browser-tasks-with-xbrowser… · home topics ai-tools article
[ARTICLE · art-15093] src=dev.to pub= topic=ai-tools verified=true sentiment=↑ positive

Automate Browser Tasks with xbrowser: A Developer's Guide to Web Automation

A developer created xbrowser, an open-source CLI tool that automates browser tasks like web searching, scraping, and crawling from the command line. The tool ships as a single npm package with a managed Chromium build, eliminating the need for separate browser downloads, WebDriver setup, or configuration files. xbrowser enables composable commands for real-world tasks such as searching multiple search engines without API keys and extracting structured content as markdown.

read6 min publishedMay 27, 2026

Browser automation has been stuck in a rut for years. The dominant tools — Selenium, Puppeteer, Playwright — are powerful, but they're built for testing, not for real-world task automation. You want to scrape a competitor's pricing page? Write a 40-line script. Need to search Google and Bing simultaneously and compare results? That's another script. Want to chain a login flow with a data extraction step? Now you're managing async state, waiting for selectors, and praying nothing times out.

I've been writing browser automation code for years, and I kept running into the same friction: too much boilerplate for tasks that should take one command. That frustration led me to xbrowser, a CLI tool designed specifically for developers and AI agents who need to get things done in a browser without writing a full test suite every time.

Let's be clear — Playwright and Selenium are excellent at what they do. If you're writing end-to-end tests for a web application, they're the right choice. But when your use case shifts from "test my app" to "interact with the web," the cracks start to show:

What I wanted was something like curl

but for interactive browser tasks — a single command that handles the complexity and gives me the result.

xbrowser is an open-source (MIT) browser automation CLI that ships as a single npm package:

npm i -g @dyyz1993/xbrowser

That's the entire installation. No separate browser download, no WebDriver setup, no configuration files. It comes with a managed Chromium build that includes CDP fingerprint protection — meaning the sites you visit can't easily detect that you're running an automated browser.

The tool is designed around composable commands that map to real-world tasks rather than low-level browser APIs. Let me walk through the core features.

Searching the web from the command line shouldn't require an API key. xbrowser handles the browser interaction for you:

xbrowser search "headless browser automation tools" --engine google --num 10

xbrowser search "headless browser automation tools" --engine bing --num 10

xbrowser search "无头浏览器自动化工具" --engine baidu --num 10

Each command returns structured results with titles, URLs, and snippets. You can pipe them into jq

for filtering, save them to a file, or feed them directly into an AI agent's context.

This is particularly useful for competitive analysis. Want to see how your brand ranks across search engines?

xbrowser search "my product name" --engine google --num 30 | jq '.results[] | select(.url | contains("myproduct.com"))'

No API keys, no rate limits to manage, no OAuth flows. Just search and get results.

The scrape

command extracts clean, structured content from any URL:

xbrowser scrape https://example.com/blog/my-article

xbrowser crawl https://example.com --depth 3 --max-pages 100

xbrowser map https://example.com

The scrape

output is markdown by default, which means it's immediately usable — paste it into a document, feed it to an LLM, or parse it with standard text tools.

crawl

follows internal links and respects depth limits, giving you a complete content snapshot of a site. map

produces a flat list of every reachable URL, which is invaluable for SEO audits.

Here's a practical example — auditing your own site's internal link structure:

xbrowser map https://mysite.com > sitemap.txt

cat sitemap.txt | while read url; do
  count=$(xbrowser scrape "$url" | grep -c "href=")
  echo "$url: $count links"
done

This is the feature that sets xbrowser apart. Instead of writing multi-step scripts, you chain operations in a single command:

xbrowser chain "goto https://news.ycombinator.com && click '.titleline > a' && scrape"

xbrowser chain "goto https://app.example.com/login \
  && fill '#email' 'user@example.com' \
  && fill '#password' 'my-password' \
  && click '#login-button' \
  && wait '#dashboard' \
  && scrape '#dashboard'"

The chain syntax reads like natural language: go to this page, click this element, fill in that field, scrape the result. It mirrors how you'd describe the task to another person.

For AI agent workflows, this is a game-changer. An agent can construct chain commands dynamically based on user intent:

User: "Go to Hacker News, click the top story, and summarize it for me"

Agent constructs:
xbrowser chain "goto https://news.ycombinator.com && click '.titleline > a:first-of-type' && scrape"

No script generation, no debugging async code, no selector management. The agent just builds a chain string and executes it.

xbrowser ships with 67+ plugins, and the SEO suite is particularly comprehensive:

xbrowser seo backlinks --domain example.com

xbrowser seo audit https://example.com/page

xbrowser search "target keyword" --engine google --num 30 --analyze

The backlink plugin crawls referring domains, checks link status, and reports on link quality metrics. The audit plugin checks meta tags, heading structure, image alt text, and other on-page factors.

For link-building workflows, you can combine search and scraping:

xbrowser search "write for us + web development" --engine google --num 20 | \
  jq -r '.results[].url' | \
  while read url; do
    xbrowser scrape "$url" | grep -i "guidelines\|submit\|contribute"
  done

Sometimes you need to automate a complex workflow that's hard to express as a chain. That's where recording comes in:

xbrowser record my-workflow



xbrowser replay my-workflow --headless

Record your workflow once in a visible browser, then replay it on a schedule or in CI. This is perfect for:

Let me be straightforward about when to use what:

Feature xbrowser Playwright Selenium
Installation
npm i -g (one step)
npm install + browser download npm install + WebDriver
CLI-first
Yes No (library-first) No (library-first)
Search helpers
Google/Bing/Baidu built-in None None
SEO plugins
67+ built-in None None
Chain syntax
goto && click && scrape
Requires script Requires script
Record/Replay
Built-in Codegen (code output) IDE plugins
Anti-detection
CDP fingerprint protection Basic stealth plugins External tools
Test framework
Not designed for this Primary use case Primary use case

The key distinction: xbrowser is for doing things on the web. Playwright and Selenium are for testing things on the web. Different goals, different tools.

If you're building an AI agent that needs to browse the web, scrape data, perform SEO analysis, or automate repetitive browser tasks, xbrowser gives you composable commands that map directly to those tasks. If you're writing integration tests for your React app, stick with Playwright.

npm i -g @dyyz1993/xbrowser
xbrowser --help
xbrowser search "hello world" --engine google

Three commands and you're up and running. The full documentation, plugin directory, and API reference are available at xbrowser.dev. The source code is on GitHub under the MIT license.

If you're building AI agents that interact with the web, or if you're tired of writing 50-line scripts for tasks that should take one command, give it a try. Contributions and plugin submissions are welcome.

xbrowser is open source under the MIT license. Install with npm i -g @dyyz1993/xbrowser. Docs and examples at xbrowser.dev.

── more in #ai-tools 4 stories · sorted by recency
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/automate-browser-tas…] indexed:0 read:6min 2026-05-27 ·