Automating a browser
Fifteen minutes. The last core skill, and the one most real RPA work needs: driving a browser. You will launch Chrome, open a page, read something off it, and close it down — and learn the one idea that makes browser automation click.
What you need
- Tutorial 6 finished
- Google Chrome installed on the machine the robot runs on
- The Robomotion Chrome extension — needed for Inspect, which is the part of this tutorial that matters most
The flow you are building
Six nodes. Build them in order — each one earns its place.
1. Launch a browser
Add an Inject node renamed Start, then Browser → Open renamed Launch Browser.
Set Browser to chrome.
Run it. Chrome opens and stays open — the flow ends but the window remains, which is worth seeing once so the last step makes sense.
2. The browser id — the one idea to take away
Launch Browser writes a browser id onto the message. Every browser node after it takes
that id as an input.
Add Browser → OpenLink renamed Open Page:
| Property | Value |
|---|---|
| Browser Id | Message → browser_id |
| URL | Custom → https://robomotion.io |
That id is how Robomotion knows which window you mean. It is why a single flow can drive two
browsers at once — keep their ids in different fields — and it is why ordering matters: a
browser node that runs before Launch Browser has no id to use.
"It says no browser." Nine times out of ten the node is missing its Browser Id, or it sits before the node that creates one.
3. Read something off the page
Add Browser → GetValue renamed Read Heading:
| Property | Value |
|---|---|
| Browser Id | Message → browser_id |
| Selector | //h1 |
| Value | Message → heading |
Run it and put a Debug node after it: the page's first heading is now on the message, and the flow can branch on it, write it to a sheet, or send it somewhere.
4. Close what you opened
Add Browser → Close renamed Close Browser — Browser Id again — and a Stop.
Leaving browsers open is the classic way to fill a robot's memory over a weekend of scheduled runs. Close what you open.
Selectors — the actual skill
Launching a browser is easy. The hard part is telling the robot which thing on the page to act on, and that is what a selector does. Robomotion uses XPath:
| Selector | Matches |
|---|---|
//h1 | the first heading |
//*[@id="search"] | the element whose id is search |
//button[text()="Submit"] | a button whose visible text is exactly Submit |
//div[@class="price"]//span | a span inside a div of class price |
You do not have to write these. With the extension installed, browser nodes have an Inspect button: press it, click the element on the page, and the selector is filled in for you. There is also a recorder that watches you do the task once and writes the nodes.
See Inspect and Record.
Inspect offers a working selector; a durable one is a judgement call. Prefer an id, a
data- attribute or visible text over a long chain of generated class names — those change at
the site's next deploy and your flow breaks on a Tuesday for no visible reason.
When you are ready for the real thing
This flow is deliberately bare. A browser automation you would actually schedule adds things this one leaves out — asking the user for input, retrying when a page fails to load, catching errors, giving up gracefully after N attempts.
The Open a Web Page template in Start with an example is exactly that flow: the same six nodes, plus a Catch, a retry loop built from Go To and Label, and a Function node that tidies whatever the user typed before it is used as a URL.
Import it and read it now that you know what the middle of it is doing. That is the shape real work takes, and none of it is new — it is tutorial 5 and tutorial 4 applied to a browser.
What you have learned
- A browser session is identified by a browser id that every node passes along
- Selectors say which element, and Inspect writes them for you
- Close what you open
Next
You have finished the seven tutorials. You can read any template in the library and change it.
Browser package reference — every node, including the ones this tutorial did not use: clicking, typing, scraping tables, switching frames and windows.