Agentic Browser
Getting Started

Getting Started

Learn how to use Agentic Browser for web automation tasks.

Launching Agentic Browser

  1. Open the Hub
  2. Click on Agentic Browser in the tool list
  3. Wait for the environment to initialize
  4. You’ll see the chat interface and browser view

Interface Overview

Chat Panel

The left side shows the conversation interface where you:

  • Type your automation requests
  • See agent responses and progress
  • Approve sensitive actions
  • View task history

Browser View

The right side shows a live view of the automated browser:

  • Watch agents navigate in real-time
  • See what elements are being clicked
  • Monitor form inputs
  • Verify actions before approval

Your First Task

Simple Navigation

Start with a basic task:

Go to wikipedia.org and search for "artificial intelligence"

Watch as Agentic Browser:

  1. Opens the browser
  2. Navigates to Wikipedia
  3. Finds the search box
  4. Types the query
  5. Submits the search

Data Extraction

Try extracting information:

Go to news.ycombinator.com and get the titles of the top 5 stories

The agent will:

  1. Navigate to Hacker News
  2. Identify the story elements
  3. Extract the titles
  4. Return them in a structured format

Understanding Agent Collaboration

Agentic Browser uses multiple specialized agents:

Orchestrator

The Orchestrator:

  • Receives your request
  • Breaks it into subtasks
  • Assigns work to other agents
  • Coordinates the overall flow

Web Surfer

The Web Surfer handles:

  • Page navigation
  • Element identification
  • Clicking and scrolling
  • Form interactions

Coder

The Coder helps with:

  • Complex data extraction
  • Custom scripts for special cases
  • Parsing structured data
  • Handling edge cases

File Surfer

The File Surfer manages:

  • Downloaded files
  • Screenshots
  • Exported data
  • Document handling

Approving Actions

When Approval Is Needed

Agentic Browser pauses for sensitive actions:

  • Login forms: Before entering credentials
  • Personal data: Before submitting forms with your info
  • Downloads: Before saving files
  • Payments: Before any financial actions

The Approval Dialog

When approval is needed, you’ll see:

  1. Action description: What the agent wants to do
  2. Details: Specific values or URLs involved
  3. Options: Approve, Modify, or Cancel

Modifying Actions

You can adjust before approving:

Agent: I'll enter "john@example.com" in the email field

You: Use my work email instead: john@company.com

Task Types

Research Tasks 

Find the pricing for the top 3 project management tools and summarize
Search for reviews of the iPhone 15 and list the common pros and cons

Data Collection 

Go to this LinkedIn company page and get their employee count and description
Extract the product names and prices from this e-commerce category page

Form Automation 

Fill out the contact form on example.com with:
- Name: John Doe
- Email: john@example.com
- Message: Interested in your services

Web Testing 

Go to our staging site and verify the login form works
Check if all navigation links on the homepage are working

Tips for Success

Be Specific

Good:

Go to amazon.com, search for "wireless headphones",
filter by 4+ stars, sort by price low to high,
and get the name and price of the first 5 results

Too vague:

Find me some headphones

Break Down Complex Tasks

Instead of one huge request, consider multiple steps:

Step 1: Navigate to the target website
Step 2: Let me see the page structure
Step 3: Now extract the data I need

Provide URLs When Possible 

Go to https://example.com/products/category/electronics
and extract all product names

Watch the Browser

The live browser view helps you:

  • Verify the agent is on the right track
  • Catch issues early
  • Understand what’s happening
  • Provide corrections if needed

Next Steps

Agent Types