Integration · Research and analyze

Cloudflare Browser Run and Meteor AI agents

Connect Cloudflare Browser Run to a Met: open pages, turn them into text, take screenshots, extract data and crawl sites with the official MCP.

Book a demo See every integration

How does Cloudflare Browser Run connect to Meteor?

With the Cloudflare Browser Run ready-made connector, your Met opens web pages in Cloudflare's browser with the account the admin authorizes. It turns them into Markdown or brings their HTML, takes screenshots at the viewport size you ask for, exports them to PDF, pulls the links or the elements you point to with a CSS selector and extracts structured JSON data from an instruction. It also starts a crawl of a whole site and then brings its results. It is the official Cloudflare server and usage runs on your Cloudflare account. Starting and cancelling a crawl and closing a browser session start switched off until the admin turns them on.

What it automates

What a Meteor Met does with Cloudflare Browser Run

Read a page someone sends in the chat

A customer or someone on the team pastes a link. The Met opens the page, turns it into text and answers with what it says, instead of asking the person to copy it.

Pull the price and name from a product page

The Met opens the product page and extracts as JSON the fields you ask for, such as name, price and availability, following the schema your team left in the skill.

Check how your page looks on mobile

The Met takes a screenshot at a phone's viewport size, checks the page's links and tells you whether the button or the price you were looking for shows up.

Crawl a whole site

If the admin turns crawling on, since it starts switched off, the Met starts a crawl to review every page of a site, waits for it to finish and brings the records to summarize what it found.

13 tools

Everything your Met can ask Cloudflare Browser Run for

Each one is a real action in the connector. You can switch off the ones you would rather it never used.

Tool	What it does
`get_url_html_content`	Brings the HTML content of the URL you give it.
`get_url_markdown`	Opens the page and converts its content to Markdown.
`get_url_screenshot`	Takes a screenshot of the page; optionally, at the viewport size you specify.
`get_url_pdf`	Renders the page into a PDF document.
`get_url_snapshot`	Returns the page HTML and a screenshot in a single call.
`scrape_url_elements`	Extracts elements from the page using a CSS selector.
`get_url_json`	Extracts structured JSON data from the page, with an instruction and a JSON schema.
`get_url_links`	Returns the list of links on the page.
`start_crawl`	Starts an asynchronous crawl of a site and returns the job identifier. It starts switched off.
`get_crawl_result`	Returns the status and records of a crawl job.
`cancel_crawl`	Cancels a running crawl. It starts switched off.
`list_browser_sessions`	Lists the account's active Browser Run browser sessions.
`kill_browser_session`	Closes a Browser Run browser session by its identifier. It starts switched off.

The limits

What the Cloudflare Browser Run integration does NOT do

We would rather say it here than have you find out during rollout.

It does not sign in to sites with a username and password or fill in forms: the tools open, read, capture and crawl pages, nothing else.
It does not ask for confirmation in the chat before using an action that is on. The admin decides in Connections which actions stay on; starting and cancelling crawls and closing browser sessions start switched off.
It does not pick an account on its own if your credentials see several: the tools work on one Cloudflare account and, with several, you have to say which.
It does not store the pages it reads: each query opens the page at that moment, so if the page changes, the answer changes.
It does not send you the screenshot or the PDF as a file by itself: the Met receives them to review them and tells you what it sees.

Before you start

What you need to connect Cloudflare Browser Run

1

A Cloudflare account

Connect with a Cloudflare account that has access to Browser Run. If that account sees several Cloudflare accounts, keep the identifier of the one you will use at hand.
2

Connect it in Met

In Met, open Connections → Cloudflare Browser Run → Connect. A Cloudflare window opens so you can sign in and authorize Met. Then you choose which actions stay on.

Frequently asked questions

Questions about Cloudflare Browser Run and Meteor

It opens public pages by their URL and reads them the way a browser would. It does not sign in to private sites, so whatever sits behind a username and password stays out.

Markdown brings the page text to read it in full. JSON pulls only the fields you ask for, such as name and price, following an instruction and a schema.

Yes, if the admin turns that action on, since it starts switched off. Meanwhile, the Met can check the crawl status with its identifier.

You do, with your own Cloudflare account and plan. Meteor does not resell Cloudflare: the connector uses your company's account.

Step by step