Let your AI assistant browse with Browser tools

Give your AI assistant a real browser. With the Zenrows MCP server connected, Claude, Cursor or any MCP client can open pages in Zenrows Browser, the browser for agents, then click, type, read, take screenshots and close the session, all from a plain-language request. The same steps are available as zenrows browser commands for scripts and terminal agents. Uses: Browser through the MCP browser tools and the zenrows browser CLI command

Ask your AI agent

With the Zenrows MCP server connected, paste any of these prompts. Each one describes the steps, the data to return and the cleanup in plain words, and your agent picks the browser tools on its own. Log in and read a protected page:

Use the Zenrows browser to log in at https://www.scrapingcourse.com/login with email
admin@example.com and password password. When the dashboard loads, tell me the first
product name, take a screenshot of the product grid, then close the browser session.

Search a site and compare results:

Use the Zenrows browser to open https://www.scrapingcourse.com/ecommerce/, search for
"hoodie", and tell me how many results there are. Open the first result and give me
its name, price and available sizes. Close the browser session when you finish.

Fill a form and confirm it worked:

Use the Zenrows browser to open
https://www.scrapingcourse.com/ecommerce/product/abominable-hoodie/, choose size M and
color Blue, set the quantity to 2, click "Add to cart", and tell me the confirmation
message. Then close the session.

Browse as a visitor from another country:

Use the Zenrows browser to open https://httpbin.io/ip as a visitor from Germany and tell me
which IP the site sees. Close the session afterwards.

How the assistant uses the tools

The assistant chains the browser tools in order. A login request, for example, runs as:

Open a session

browser_navigate opens the URL in a new Zenrows Browser session and returns a session_id. It also takes proxy_country or proxy_region.

Act on the page

browser_fill types the email and password, and browser_press_key presses Enter. browser_click, browser_select_option and browser_scroll cover the rest of a flow.

Wait and read

browser_wait_for_selector waits for the dashboard, then browser_get_text, browser_get_cookies or browser_screenshot return what you asked for.

browser_close ends the session and frees its concurrency slot.

Ask for the close step in your prompt: it ends billing as soon as the task is done.

Run the same flow from the terminal

The CLI exposes each step as a command. open prints a session ID that every other command takes with --session.

CLI

# Open a session and keep its ID
SESSION=$(zenrows browser open https://www.scrapingcourse.com/ecommerce/ --json | jq -r .session_id)

# Search the catalog
zenrows browser fill --session $SESSION --selector "#woocommerce-product-search-field-0" --value "hoodie"
zenrows browser press-key --session $SESSION --key Enter
zenrows browser wait-for-selector --session $SESSION --selector "li.product"
zenrows browser get-text --session $SESSION --selector ".woocommerce-result-count"

# Open the first result, read it and capture it
zenrows browser click --session $SESSION --selector "li.product a"
zenrows browser get-text --session $SESSION --selector "h1.product_title"
zenrows browser screenshot --session $SESSION --out product.png

# Always close the session
zenrows browser close --session $SESSION

Add --json to any command for structured output, for example zenrows browser open <url> --json returns session_id, title, finalUrl and expires_at, which is how the script above reads the session ID with jq. Install the CLI with npm install -g @zenrows/cli, see the CLI guide.

Example output

The MCP login flow, tool by tool:

browser_navigate         → {"session_id": "8562d040-...", "url": "https://www.scrapingcourse.com/login",
                            "title": "Login Challenge to Learn Web Scraping - ScrapingCourse.com"}
browser_fill (#email)    → {"ok": true}
browser_fill (#password) → {"ok": true}
browser_press_key Enter  → {"ok": true}
browser_wait_for_selector #product-grid → {"ok": true}
browser_get_text .product-name → Chaz Kangeroo Hoodie
browser_get_cookies      → AWSALBCORS, _ga, AWSALB, scrapingcoursecom_session, _ga_NZGD14H87G, XSRF-TOKEN
browser_close            → {"ok": true}

The CLI flow (trimmed):

✓ Session bb15ec60-a544-494f-8328-1ae88ffdbe4d · "Ecommerce Test Site to Learn Web Scraping – ScrapingCourse.com" · expires 2026-10-14T23:49:48Z · run r-d5f24dc7
Showing 1–16 of 19 results
Eos V-Neck Hoodie
✓ Wrote 109647 bytes → product.png
✓ Closed session bb15ec60-a544-494f-8328-1ae88ffdbe4d.

Customize it

Change	How
Browse from a country	Ask for “proxy country US”, or `zenrows browser open <url> --proxy-country us`
Browse from a world region	Ask for “proxy region EU”, or `--proxy-region eu`
Save the page as PDF	Ask for a PDF (`browser_generate_pdf`), or `zenrows browser pdf --session $SESSION --out page.pdf`
Run JavaScript on the page	`browser_evaluate`, or `zenrows browser evaluate --session $SESSION --script "document.title"`
Reuse a login	`browser_get_cookies` then `browser_set_cookies` in the next session, or `zenrows browser cookies` / `set-cookies`
Drive the browser from your own code	Print the WebSocket URL with `zenrows browser connect`, see Connect Playwright or Puppeteer
Read a page without a browser	Ask the assistant to “read the page with Zenrows” (the `scrape` tool), see Turn any web page into LLM-ready markdown

Troubleshooting

The assistant does not offer browser tools: the MCP server is not connected in this client. Follow Connect your AI agent to Zenrows and restart the client.
A step fails with “element not found”: the page was still loading. Ask the assistant to wait for the element first (browser_wait_for_selector), or run wait-for-selector in the CLI.
zenrows browser open reports a navigation error: the target URL didn’t load. Close the session with the ID in the error, check the URL, and run open again.
Sessions stay open after a conversation: the prompt did not ask to close them. End prompts with “close the browser session when you finish”. Sessions the CLI and MCP tools open also end on their own at the expires_at time that zenrows browser open --json returns.

Connect your AI agent to Zenrows

Set up the MCP server in Claude, Cursor or VS Code.

Prompts that put Zenrows MCP to work

Ready-made prompts for Fetch, Extract, Crawl, Batch and Browser.

Build a browser agent with Zenrows

Let Claude or GPT drive the browser with the Zenrows tools.

Automate scraping from your terminal with the CLI

Chain fetch, extract, batch and browser commands in scripts.

Start Here

Foundations Course

Fetch Recipes

Extract Recipes

Crawl Recipes

Batch Recipes

Browser Recipes

Pipelines

Use Cases

By Surface

Production

Let your AI assistant browse with Browser tools

Ask your AI agent

How the assistant uses the tools

Run the same flow from the terminal

Example output

Customize it

Troubleshooting

Connect your AI agent to Zenrows

Prompts that put Zenrows MCP to work

Build a browser agent with Zenrows

Automate scraping from your terminal with the CLI

​Ask your AI agent

​How the assistant uses the tools

​Run the same flow from the terminal

​Example output

​Customize it

​Troubleshooting

​Related recipes

Connect your AI agent to Zenrows

Prompts that put Zenrows MCP to work

Build a browser agent with Zenrows

Automate scraping from your terminal with the CLI

Ask your AI agent

How the assistant uses the tools

Run the same flow from the terminal

Example output

Customize it

Troubleshooting

Related recipes