Skip to main content
Common patterns for browser automation. These examples show the agent’s workflow — you just describe what you need in plain language, and the agent handles the tool calls.

​
Login and Navigate

The agent can log into web apps using your credentials and navigate authenticated pages. 
"Go to app.example.com, log in with my credentials, and navigate to the Reports page"
What happens under the hood:
1

Open and navigate

The agent opens the browser and navigates to the login page.
2

Snapshot the page

Takes an accessibility snapshot to find the email/password inputs and login button — each gets a reference like @e3, @e5, @e7.
3

Fill and submit

Types your email and password into the input fields, then clicks the login button.
4

Wait and continue

Waits for the page to load after login, then navigates to the target page.
If you’re already logged into a site in the browser, the agent can reuse that session — no need to log in again.

​
Fill a Complex Form

The agent can handle multi-field forms with dropdowns, checkboxes, and text areas. 
"Fill out the support ticket form on our internal tool — set priority to High,
category to Billing, and describe the issue as 'Customer charged twice for subscription'"
The agent snapshots the form, identifies each field by its label, fills text inputs, selects dropdown values, and clicks submit.

​
Extract Data from a Page

Use JavaScript execution to pull structured data from any webpage. 
"Go to our analytics dashboard and extract the top 10 metrics from the summary table"
What the agent does: 
// The agent runs JavaScript directly in the page context
Array.from(document.querySelectorAll('table.summary tr'))
  .slice(0, 10)
  .map(row => ({
    metric: row.cells[0].textContent,
    value: row.cells[1].textContent
  }))
This returns structured data that the agent can format as a table, export to a spreadsheet, or use for further analysis.

​
Monitor Network Traffic

See what API calls a page is making — useful for debugging or discovering internal endpoints. 
"Open the dashboard, click the refresh button, and show me what API calls it makes"
The agent uses network inspection to capture all HTTP requests the page triggers, showing URLs, methods, status codes, and response sizes. This is especially powerful for discovering internal APIs.

​
Take Targeted Screenshots

Capture specific elements or regions instead of the full page. 
"Take a screenshot of just the revenue chart on the dashboard"
The agent can target screenshots by:
  • Element reference — snapshot the page, find the chart’s ref, capture just that element
  • CSS selector — target elements like div[data-testid="revenue-chart"]
  • Coordinates — capture a specific pixel region

​
Send Keyboard Shortcuts

Trigger app-specific keyboard shortcuts for power-user workflows. 
"Open the command palette with Cmd+K and search for 'billing'"
The agent can send any key combination including modifier keys (Shift, Control, Alt, Meta/Cmd).

​
Multi-Step Workflows

Combine multiple browser actions into complex workflows. 
"Go to our HR portal, check each team member's profile, and compile their
job titles and departments into a spreadsheet"
For tasks like this that involve iterating through many items, the agent often discovers a more efficient approach — see API Discovery for how the agent can find internal APIs and fetch all data in parallel instead of clicking through one by one.
  • Be specific about what you want — “Extract the employee names and emails from the table” is better than “get the data”
  • Mention if you’re already logged in — saves time skipping the login flow
  • Describe the page structure if it’s complex — “the data is in the second tab, under the Summary section”
  • Ask for a specific output format — “put it in a spreadsheet” or “format as a table”
  • Page not loading? — The agent will retry navigation. If it keeps failing, check that the URL is correct and the site is accessible.
  • Can’t find an element? — The agent re-snapshots the page after navigation. If elements load dynamically, it may need to wait or scroll first.
  • Login not working? — Some sites use CAPTCHAs or multi-factor auth that the browser can’t automate. You may need to log in manually first, then let the agent continue.
  • Interactions seem flaky? — The agent will re-snapshot and retry. Dynamic pages with animations may need a brief wait between actions.