Skip to main content
Not sure which browser automation approach to use? This guide will help you make the right choice.

​
Quick Decision Tree

1

Is it a simple 1-2 action task?

Yes → Use Browser ActionsNo → Continue to next step
2

Is the primary goal data extraction?

Yes → Use Browser CodeNo → Continue to next step
3

Does it require multi-step navigation or complex workflow?

Yes → Use Browser AgentsNo → Use Browser Actions

​
By Use Case

​
Simple Interactions

Example: “Click the login button and enter email”

Browser Actions

Best Choice - Fast, cheap, and perfect for simple tasks
 Why?
  • Completes in 1-2 steps
  • Cacheable for cost savings
  • Fast execution
  • Predictable outcome

​
Data Extraction

Example: “Extract all product names and prices from the page”

Browser Code

Best Choice - Precise, reliable extraction with structured output
 Why?
  • Returns structured data
  • Precise control over extraction logic
  • Cacheable code generation
  • Reliable and repeatable
Alternative: Use browser_assess_page for simpler extraction with schema validation.

​
Multi-Page Workflows

Example: “Complete the checkout process from cart to confirmation”

Browser Agents

Best Choice - Autonomous navigation with unlimited steps
 Why?
  • No step limit
  • Handles complex flows
  • State management
  • Adaptive to page changes

​
Visual Understanding

Example: “Find and click the green ‘Confirm’ button in the modal”

Computer Use Agent

Best Choice - Vision-based understanding of layout
 Why?
  • Understands visual layout
  • Extended reasoning capability
  • Works with complex visual interfaces
  • Can identify elements by appearance

​
Comparison Matrix

​
By Task Complexity

Complexity LevelTask ExamplesBest ChoiceAlternative
Low
(1-2 steps)		• Click button
• Fill single field
• Navigate to page		Browser Actions-
Medium
(3-10 steps)		• Multi-step form
• Login flow
• Search and filter		Browser AgentsBrowser Actions (if predictable)
High
(10+ steps)		• Complete checkout
• Multi-page wizard
• Complex workflow		Browser Agents-
Data Focused• Extract table data
• Scrape listings
• Parse structured info		Browser Codebrowser_assess_page

​
By Priority

PriorityRecommendationWhy
Cost OptimizationBrowser Actions (with caching)5 credits per cached execution
ReliabilityBrowser CodeExplicit, deterministic logic
FlexibilityBrowser AgentsHandles unpredictable scenarios
SpeedBrowser ActionsFast execution, minimal overhead
PrecisionBrowser CodeExact control over operations

​
By Output Needs

What You NeedBest ChoiceExample
No output (just actions)Browser Actions or AgentsClick, navigate, type
Simple confirmationBrowser Actions”Button clicked successfully”
Structured dataBrowser Code[{name: "...", price: "..."}]
Page assessmentbrowser_assess_pageEvaluated structured data
ScreenshotsAny + browser_get_screenshotPNG image data

​
Common Scenarios

Task: Search for “wireless headphones”, filter by price under $100, and add first result to cart 
{
  "type": "browser_action",
  "action": "Search for 'wireless headphones' and click the first result under $100"
}
// ⚠️ Might hit 2-step limit
Verdict: Use Browser Agents - requires multiple steps (search, filter, select, add to cart)

​
Scenario 2: Extract Product Listings

Task: Extract all products with name, price, and rating from the current page 
{
  "type": "execute_javascript",
  "prompt": "Extract all products with name, price, and rating as array of objects",
  "return_by_value": true
}
// ✅ Precise, structured output
Verdict: Use Browser Code for complex extraction or browser_assess_page for simpler needs

​
Scenario 3: Login and Navigate

Task: Log in to dashboard and navigate to settings page 
{
  "type": "browser_action",
  "action": "Click login button and navigate to settings"
}
// ⚠️ Only if login is 1 click
Verdict: Use Browser Agents - login flows often require multiple steps

​
Scenario 4: Click Submit Button

Task: Scroll to the bottom and click the submit button 
{
  "type": "browser_action",
  "action": "Scroll to bottom and click submit button"
}
// ✅ Perfect fit - 2 simple actions
Verdict: Use Browser Actions - simple, predictable, cacheable

​
Cost Considerations

​
Optimize for Cost

  1. Use Browser Actions for repeated simple tasks
    • Fresh: 50 credits
    • Cached: 5 credits (10x cheaper!)
  2. Cache Browser Code for repeated extractions
    • Code generation cached
    • Subsequent runs cheaper
  3. Avoid over-using Agents for simple tasks
    • Variable cost based on steps
    • More expensive than Actions

​
Example Cost Comparison

Task: Click login button (repeated 100 times)
ApproachCost
Browser Actions (cached)5 credits × 100 = 500 credits
Browser Actions (fresh)50 credits × 100 = 5,000 credits
Browser Agents~2-5 steps × cost/step × 100 = Higher
For repeated tasks, Browser Actions with caching provides massive cost savings!

​
Still Not Sure?

​
Start Simple, Upgrade as Needed

  1. Try Browser Actions first if the task seems simple
  2. Upgrade to Agents if you hit the 2-step limit
  3. Switch to Code if you need precise data extraction

​
Ask Yourself

  • 1-2 steps: Browser Actions
  • 3+ steps: Browser Agents
  • Just extraction: Browser Code
  • Yes, frequently: Browser Actions (for caching)
  • No, one-time: Any approach works
  • Varies by data: Browser Code (caches generation)
  • Very predictable: Browser Actions or Code
  • Some variation: Browser Agents
  • Highly dynamic: Browser Agents
  • None (just actions): Browser Actions
  • Structured data: Browser Code
  • Completion confirmation: Browser Agents

​
Quick Reference

Browser Actions

When: Simple, fast, repeatedCost: 5-50 creditsLimit: 2 steps

Browser Agents

When: Complex, multi-step, adaptiveCost: VariableLimit: None

Browser Code

When: Data extraction, precise controlCost: MediumLimit: Logic-based

​
Need Help?

For detailed implementation, see individual guides or contact support at [email protected].