Skip to main content

​
Executing actions

Stagehand has an act() function that can be used to execute actions on a page using natural language. Here’s an example of Stagehand to find jobs on LinkedIn: Take actions with Stagehand This workflow is as simple as the following lines of code: 
const page = stagehand.page
// Navigate to the URL
await page.goto("https://linkedin.com");
// Click on jobs menu selection
await page.act("click 'jobs'");
// Click on first posting
await page.act("click the first job posting");
The page object extends the Playwright page object, so you can use any of the Playwright page methods with it.

​
Get structured data from the page

You can use extract() to get structured data from the page. Here’s an example of how to extract the job title from the job posting: 
const { jobTitle } = await page.extract({
	instruction: "Extract the job title from the job posting",
	schema: z.object({
		jobTitle: z.string(),
	}),
});
Stagehand uses Zod for TypeScript and Pydantic for Python to help you define the schema of the data to be extracted.

​
Preview/Cache an action

Sometimes you want to preview an action before it’s executed. You can do this by calling page.observe() before act(). 
const [action] = await page.observe("click 'jobs'");

console.log("Going to execute action:", action)
// You can add extra logic here to validate/cache the action

await page.act(action)
action will be a JSON object that describes the action to be taken. 
// action is a JSON-ified version of a Playwright action:
{
	description: "The jobs link",
	method: "click",
	selector: "/html/body/div[1]/div[1]/a",
	arguments: [],
}
For more on caching, see the caching docs.

​
What actions can I take?

Stagehand maps natural language to Playwright actions. We generally support the following actions:
ActionDescription
scrollIntoViewScrolls an element into the visible area of the browser window
scrollToScrolls to a specific percentage of the page height
fillFills in form fields with specified text
typeTypes text into input fields (alias for fill)
pressSimulates pressing keyboard keys
clickClicks on elements matching the specified selector
nextChunkScrolls the height of the viewport by 100%
prevChunkScrolls the height of the viewport by -100%
Each of these actions can be triggered using natural language commands. For example: 
await page.act("scroll to the bottom of the page")
await page.act("fill in the username field with 'john_doe'")
await page.act("scroll the modal to the next chunk")