Interaction Actions

Actions for clicking, typing, scrolling, and interacting with page elements

Interaction actions allow the agent to manipulate page elements like a human user.

click

Click on an element by its index from the browser state.

index · int · required

Element index from browser_state. Must be ≥ 1.

Element at index 0 represents the entire page and cannot be clicked.

coordinate_x · int

Horizontal coordinate relative to viewport left edge (optional, for coordinate-based clicking)

coordinate_y · int

Vertical coordinate relative to viewport top edge (optional, for coordinate-based clicking)

Example

from browser_use import Agent, Browser, ChatBrowserUse

agent = Agent(
    task="Go to example.com and click the first link",
    llm=ChatBrowserUse(),
    browser=Browser()
)

await agent.run()

The agent automatically detects new tabs opened by clicks and provides information about them.

Implementation: browser_use.tools.service:566 (index-based), browser_use.tools.service:521 (coordinate-based)

---

input

Type text into an input field.

index · int · required

Element index from browser_state pointing to an input field

text · string · required

The text to type into the field

clear · bool · default: "True"

Whether to clear the field before typing. Set to False to append text.

Example

from browser_use import Agent, Browser, ChatBrowserUse

agent = Agent(
    task="Go to google.com and search for 'browser use'",
    llm=ChatBrowserUse(),
    browser=Browser()
)

await agent.run()

For autocomplete fields, wait for suggestions to appear before selecting. The action will add a brief delay for JavaScript-driven autocomplete.

Implementation: browser_use.tools.service:639

---

scroll

Scroll the page or a specific element up or down.

down · bool · default: "True" · required

Direction to scroll. True = scroll down, False = scroll up

pages · float · default: "1.0"

Number of pages to scroll.

• 0.5 = half page
• 1.0 = full page
• 10.0 = scroll to bottom/top

Range: 0.5 to 10.0

index · int

Optional element index to scroll within a specific element (e.g., dropdowns, custom scrollable containers)

Example

from browser_use import Agent, Browser, ChatBrowserUse

agent = Agent(
    task="Go to hackernews and scroll down 3 pages",
    llm=ChatBrowserUse(),
    browser=Browser()
)

await agent.run()

Viewport height is automatically detected for accurate scrolling. Multi-page scrolls execute sequentially to ensure each completes.

Implementation: browser_use.tools.service:1241

---

send_keys

Send special keyboard keys or shortcuts to the page.

keys · string · required

Keys to send. Can be:

• Special keys: Escape, Enter, Tab, PageDown, PageUp, ArrowDown, ArrowUp, ArrowLeft, ArrowRight
• Key combinations: Control+c, Control+v, Control+a
• Or any other keyboard input

Example

from browser_use import Agent, Browser, ChatBrowserUse

agent = Agent(
    task="Press Tab three times then Enter",
    llm=ChatBrowserUse(),
    browser=Browser()
)

await agent.run()

Use send_keys for keyboard navigation when buttons can’t be clicked or for form submission shortcuts.

Implementation: browser_use.tools.service:1346

---

Related Actions

• Navigation Actions - Navigate between pages
• File Actions - Upload and manage files