Skip to main content
Browser Actions enable you to interact with webpages before data extraction. Choose between two modes based on your needs: deterministic control or AI-powered flexibility. Common uses:
  • Load dynamic content: Trigger lazy-loaded data, infinite scrolls, and pagination
  • Submit forms: Enter search terms, apply filters, or configure settings
  • Navigate pages: Click buttons, open dropdowns, and interact with UI elements
  • Capture state: Take screenshots, collect cookies, or record network traffic

Two modes

Browser Actions offers two approaches to automate webpage interactions:
  1. Instructions Mode
  2. LLM Mode

Instructions Mode

Manually define each action step with full control over execution sequence. Best for:
  • Predictable, repeatable automation
  • Cost-sensitive applications (lower pricing)
  • Static pages with consistent selectors
  • Fast execution without LLM overhead
How it works:
  1. Define actions in the browser_actions as array of actions
  2. Specify selectors, delays, and sequences
  3. Nimble executes steps exactly as configured
Pricing: Based on driver usage (vx6, vx8, vx10, etc.) - no token costs

Instructions Mode

Full control with predefined action steps

LLM Mode

Describe desired interactions in natural language and let AI determine the best execution strategy. Best for:
  • Zero-setup automation
  • Pages that change frequently
  • Complex interaction sequences
  • Intelligent decision-making
How it works:
  1. Define actions in the browser_actions as a natural language prompt (e.g., “scroll until all items load, open filters, click ‘On Sale’”)
  2. Nimble’s AI agent takes control of the browser in real-time
  3. AI adapts to page variations and figures out the optimal sequence
Pricing: Uses vx14 driver + token consumption (higher cost than deterministic)

LLM Mode

AI-powered automation with natural language prompts

Quick comparison

FeatureInstructions ModeLLM Mode
SetupManual step configurationNatural language prompt
CostLower (driver only)Higher (driver + tokens)
FlexibilityFixed sequenceAdapts dynamically
MaintenanceUpdate when page changesSelf-healing
SpeedFaster (no inference)Slower (LLM processing)
ControlFull controlAI-determined
Best forStable pages, cost controlChanging pages, zero setup
Both modes require page rendering to be enabled (render: true) and operate within a global 120-second timeout.

Choosing the right mode

Use Instructions Mode when:
  • You need low-cost, predictable automation
  • Page structure is stable and selectors don’t change often
  • You want full control over every action
  • Performance is critical
Use LLM Mode when:
  • You want zero-setup, hands-free automation
  • Pages change frequently and maintaining selectors is challenging
  • You need intelligent handling of dynamic elements
  • Higher cost is acceptable for reduced maintenance

Usage examples

Instructions Mode

Define exact steps to execute:
from nimble import Nimble

nimble = Nimble(api_key="YOUR-API-KEY")

result = nimble.extract({
    "url": "https://www.example.com",
    "render": True,
    "browser_actions": [
        {
            "wait_and_type": {
                "selector": "input[type='search']",
                "value": "laptop"
            }
        },
        {
            "wait_and_click": {
                "selector": "button[type='submit']"
            }
        }
    ]
})

print(result)

LLM Mode

Let AI figure out the steps:
from nimble import Nimble

nimble = Nimble(api_key="YOUR-API-KEY")

result = nimble.extract({
    "url": "https://www.example.com",
    "render": True,
    "browser_actions_prompt": "Search for 'laptop' and apply filters for 'On Sale' items"
})

print(result)

Next steps