Prompts that put Zenrows MCP to work

With the Zenrows MCP server connected, your AI agent picks the right tool from a plain-English request. You never need parameter names: describe the page, the outcome and the format, and the agent chooses the tool and its settings, including JavaScript rendering, premium proxies and the country to browse from. Copy a prompt, swap the URL, and run it. Uses: Zenrows MCP server: scrape, extract, crawl_*, batch_*, browser_*, account_usage (see the tool reference)

Connect the server

Connect the Zenrows MCP server to your client first: see Connect your AI agent to Zenrows. Every connection option exposes the same tools.

Read and summarize pages

Ask for any page in plain words. Zenrows renders JavaScript and handles anti-bot protection on its own, and returns clean markdown your agent can read.

Outcome	Prompt
A summary of one page	`Use Zenrows to read https://www.scrapingcourse.com/ecommerce/product/abominable-hoodie/ and summarize it in three bullets.`
Answers from a JavaScript-heavy page	`Use Zenrows to read https://www.scrapingcourse.com/javascript-rendering and list every product name and price on it.`
The page as a specific country sees it	`Use Zenrows to read <product url> as a visitor from the UK and tell me the price and currency shown.`
A screenshot to check the layout	`Use Zenrows to take a full-page screenshot of https://www.scrapingcourse.com/ecommerce/ and describe the layout.`
Links from a page	`Use Zenrows to get every link on https://www.scrapingcourse.com/ecommerce/ and list the product pages among them.`

Get structured fields

Ask for the fields you want. Zenrows returns them as JSON instead of a full page, which keeps your agent’s context small.

Outcome	Prompt
Product fields as JSON	`Use Zenrows to get the title, current price and stock status of https://www.amazon.com/dp/B08Z6X4NK3 as a shopper in the US sees it.`
Exactly the elements you name	`Use Zenrows to read https://www.scrapingcourse.com/ecommerce/ and give me a table of every product's name and price, using the CSS selectors .product-name and .price.`
A comparison across retailers	`Use Zenrows to get the price of each of these products and compare them in one table: <url1> <url2> <url3>`

Discover every page on a site

Give a start page and say which pages you want. The agent crawls the site and returns the URLs, or the pages themselves.

Outcome	Prompt
Every product URL in a catalog	`Use Zenrows to find every product page on https://www.scrapingcourse.com/ecommerce/, spending at most 20 page loads, and list the URLs.`
A site map of a docs section	`Use Zenrows to map every page under https://docs.example.com/guides/ and group the results by path.`
The pages themselves, not only URLs	`Use Zenrows to crawl <url>, keep the first 10 pages with their content, and summarize the first three.`

Scrape at scale

Hand the agent the whole list, a hundred URLs or a hundred thousand. It sends them to Zenrows as one job that runs in parallel in the cloud, so the conversation stays short and the results wait for you.

Outcome	Prompt
A whole URL list as one job	`Use Zenrows to scrape every URL in urls.txt as one batch job, wait for it to finish, and save each page as markdown in a folder called pages/.`
Structured data for a full catalog	`Use Zenrows to find every product page on https://www.scrapingcourse.com/ecommerce/, scrape them all in one batch, and give me name, price and SKU for each as a CSV.`
A big job you check on later	`Submit these 5,000 URLs to Zenrows as one batch job and give me the job ID. Don't wait for it.` then later `Check my Zenrows batch job <job ID> and list any URLs that need a retry.`
Spend before and after	`Check my Zenrows credits, run the batch, then tell me how many credits the job used.`

A two-URL test job came back "status": "completed" with "stats": {"total": 2, "successful": 2, "spend": {"credits": 2}}.

Click, type and log in

Describe the steps the way you would to a colleague. The agent drives Zenrows Browser through them and reports back.

Outcome	Prompt
Content behind a “Load more” button	`Use the Zenrows browser to open https://www.scrapingcourse.com/button-click, click "Load more" twice, then list every product name. Close the session when done.`
A logged-in page	`Use the Zenrows browser to log in at https://www.scrapingcourse.com/login with admin@example.com and password, and tell me the heading of the page you land on. Close the session.`
A PDF of a rendered page	`Use the Zenrows browser to open <url>, wait for the main content to load, save the page as a PDF, and close the session.`
A page from a region	`Use the Zenrows browser to open <url> as a visitor from Europe and tell me which currency the prices use. Close the session.`

End every browser prompt with “close the session”. It frees the session and its concurrency slot right away.

Combine primitives in one prompt

The agent chains tools on its own when the prompt names the full outcome.

Catalog to dataset

Use Zenrows to find every product page on https://www.scrapingcourse.com/ecommerce/,
scrape them all in one batch, pull name, price and SKU from each,
and give me the result as a CSV table.

Read first, browse when needed

Use Zenrows to read <url>. If the reviews section is missing,
open the page in the Zenrows browser, click "Show all reviews", read the reviews,
and close the session.

Research brief

Use Zenrows to search Bing for "web scraping best practices",
read the top three result pages, and write a one-page brief with a source link per point.

Budget-aware run

Check my Zenrows credits. If I have more than 1,000 left,
crawl <url> for up to 50 pages and summarize them. Otherwise tell me what it would need.

Use the built-in MCP prompts

The server also ships three ready-made prompts. Clients that support MCP prompts list them in their prompt or slash-command menu, for example /mcp__zenrows__scrape_and_summarize in Claude Code.

Prompt	Arguments	What it does
`scrape_and_summarize`	`url`	Scrapes the page and returns a concise summary with key points, headings and data
`extract_structured_data`	`url`, `fields` (a JSON map of field names to CSS selectors)	Runs `extract` in `css` mode and returns clean JSON
`scrape_js_page`	`url`	Scrapes a JavaScript-rendered page with `mode=auto` and returns the rendered content as markdown

Write prompts the agent follows

Start with “Use Zenrows” so the agent reaches for live data instead of its memory.
Describe the result, not the settings: “a table of name and price”, “as a visitor from Germany”, “after the reviews load”. The agent picks the parameters.
Say the size of the job: “every product page”, “these 5,000 URLs”, “up to 50 pages”. Large lists run as one batch job.
Name elements when you know them: a CSS selector such as .price turns a broad request into a precise one.
Close browser sessions in the same prompt.

Troubleshooting

The agent answers without calling Zenrows: it used its own knowledge. Start the prompt with “Use Zenrows to…”.
The tool list is empty: the server didn’t authenticate. Check the URL is exactly https://mcp.zenrows.com/mcp and the Bearer token holds your key, then restart the client.
The agent reports a crawl or batch job as still running: large jobs take longer than one check. Ask it to keep checking until the job is complete.
AUTH004 or AUTH014: the plan or the key needs more credits. Ask the agent to check your Zenrows credits, then see Track usage and control spend.

Connect your AI agent to Zenrows

Set up the MCP server in your client, step by step.

Automate scraping from your terminal with the CLI

The same jobs as shell commands and cron.

Build a research agent with live web data

Search, read and cite sources with one agent.

Let your AI assistant browse with Browser tools

Drive Zenrows Browser from a chat.

Start Here

Foundations Course

Fetch Recipes

Extract Recipes

Crawl Recipes

Batch Recipes

Browser Recipes

Pipelines

Use Cases

By Surface

Production

Prompts that put Zenrows MCP to work

Connect the server

Read and summarize pages

Get structured fields

Discover every page on a site

Scrape at scale

Click, type and log in

Combine primitives in one prompt

Use the built-in MCP prompts

Write prompts the agent follows

Troubleshooting

Connect your AI agent to Zenrows

Automate scraping from your terminal with the CLI

Build a research agent with live web data

Let your AI assistant browse with Browser tools

​Connect the server

​Read and summarize pages

​Get structured fields

​Discover every page on a site

​Scrape at scale

​Click, type and log in

​Combine primitives in one prompt

​Use the built-in MCP prompts

​Write prompts the agent follows

​Troubleshooting

​Related recipes

Connect your AI agent to Zenrows

Automate scraping from your terminal with the CLI

Build a research agent with live web data

Let your AI assistant browse with Browser tools

Connect the server

Read and summarize pages

Get structured fields

Discover every page on a site

Scrape at scale

Click, type and log in

Combine primitives in one prompt

Use the built-in MCP prompts

Write prompts the agent follows

Troubleshooting

Related recipes