Skip to main content
LLM mode enables fully hands-free browser automation. Describe your desired interactions in natural language, and Nimble’s AI agent takes control of the browser in real-time to execute them. The AI agent adapts to page variations, handles dynamic elements intelligently, and figures out complex interaction sequences automatically.

​
When to use

Use LLM mode when you need:
  • Zero setup: No selector configuration or step definition required
  • Self-healing: Automatically adapts when pages change
  • Complex interactions: AI handles multi-step sequences intelligently
  • Dynamic pages: Works with pages that change frequently
LLM mode uses the vx14 driver and incurs additional token consumption costs based on prompt and response length. This is more expensive than deterministic flows but requires zero maintenance.

​
How it works

  1. Send a prompt: Describe what you want to accomplish in natural language
  2. AI takes control: The LLM agent analyzes the page and determines the best action sequence
  3. Real-time execution: Actions are performed live in the browser, not pre-generated
  4. Adaptive behavior: AI adjusts to page variations and handles unexpected elements
The LLM agent executes actions in real-time based on the current page state. It’s not translating your prompt into predefined steps—it’s actually controlling the browser dynamically.

​
Supported parameters

Available in - Extract.
ParameterTypeDescriptionDefault
renderBooleanEnable or disable JS rendering (required to be true)false
browser_actions_promptStringDescribed the required page interactions with natural language prompt-

​
Usage

​
Basic example

Let the AI handle all interactions: 
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 filter by 'On Sale' items"
})

print(result)

​
Infinite scroll

Handle dynamic content loading: 
from nimble import Nimble

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

result = nimble.extract({
    "url": "https://www.example.com/products",
    "render": True,
    "browser_actions_prompt": "Scroll down until all products are loaded"
})

print(result)

​
Complex interaction

Multi-step workflows: 
from nimble import Nimble

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

result = nimble.extract({
    "url": "https://www.example.com",
    "render": True,
    "browser_actions_prompt": "Click the filters button, set price range to $500-$1000, select 4-star rating and above, then apply filters"
})

print(result)

​
Example response

When the LLM agent completes your prompt, you receive the final page state and execution details. The response includes:
  • data: All related extacted data
    • data.html: Final DOM state after AI execution
  • llm_execution: Details about actions the AI performed
  • actions_performed: Step-by-step log of what the agent did
  • metadata: Execution details including task id, driver used, execution time and more
    • metadata.browser_actions: The browser actions results per step
{
  "status": "success",
  "data": {
	"html":"<!DOCTYPE html><html>...</html>"
   },
  "metadata": {
	"task_id":".....",
	"country":"US",
    "driver": "vx14",
    "execution_time_ms": 2100,
	"browser_actions": {
      "prompt": "Search for 'laptop' and filter by 'On Sale' items",
      "actions_performed": [
        "Located search input field",
        "Typed 'laptop' into search",
        "Clicked search button",
        "Waited for results to load",
        "Opened filters panel",
        "Selected 'On Sale' checkbox",
        "Applied filters"
      ],
      "success": true,
      "execution_time_ms": 5200
    }
  }
}

​
Best practices

​
Writing effective prompts

Be specific but flexible: 
✅ "Search for 'wireless headphones' and filter by price under $100"
❌ "Find some headphones"
Describe outcomes, not selectors: 
✅ "Click the 'Add to Cart' button"
❌ "Click button.btn-primary.add-cart"
Chain actions logically: 
✅ "Sort by price ascending, then scroll to load all items"
❌ "Do sorting and scrolling stuff"

​
Pricing

LLM mode costs include:
  • vx14 driver usage: Higher tier driver for AI capabilities
  • Token consumption: Based on prompt length and AI response
  • API call: Standard request fee
This is typically 3-5x more expensive than deterministic flows, but eliminates maintenance costs when pages change.