Task Modes: Scrape vs Agent vs Headful
Doppelganger provides three task modes to cover a wide range of automation scenarios. Each mode is designed for different levels of interaction and complexity.
1. Scrape Tasks#
Scrape tasks are designed for simple, fast data extraction from a single URL or a list of URLs.
Key Features#
- Minimal inputs: Only requires
URLand optionalwaittime. - Fast execution: Runs quickly without additional blocks.
- Optional extraction scripts: JavaScript can be added to process DOM or Shadow DOM content.
- Local data storage: Data is stored on your machine for privacy.
Ideal Use Cases#
- Pulling product listings or prices
- Monitoring web pages for changes
- Extracting static content from sites
2. Agent Tasks#
Agent tasks enable multi-step, programmatic automation using blocks and simulated human behaviors.
Key Features#
- Blocks-based workflow: Navigate, Click, Type, Press Keys, Run JS
- Human-like behaviors: Typing delays, scroll patterns, optional typos
- Dynamic input: Supports variables for parameterization
- Optional extraction scripts: Clean or structure complex data into JSON
Ideal Use Cases#
- Multi-step form filling and workflows
- Automated browsing with authentication
- Complex automation requiring interaction with multiple page elements
3. Headful Tasks#
Headful tasks are manual, interactive browser sessions where a human operates the browser directly. They are not agentic and do not run automated blocks.
Key Features#
- Visible browser: Tasks run in a fully visible browser window
- Human interaction only: No automation blocks executed
- Supports inspection and debugging: Run extraction scripts manually if needed
Ideal Use Cases#
- Manually exploring websites while capturing data
- Debugging and testing automation scripts
- Interacting with sites that require direct human input
Comparison Table#
| Mode | Inputs | Blocks | Human-like Behavior | Extraction Scripts | Visibility | Ideal For |
|---|---|---|---|---|---|---|
| Scrape | URL + wait | No | No | Optional | Headless | Fast, single-page data extraction |
| Agent | URL + wait + blocks | Yes | Yes | Optional | Headless | Multi-step automation, dynamic workflows |
| Headful | URL | No | Human only | Not supported | Visible | Debugging, logins, human-operated interaction, exploration |
Best Practices#
- Start with Scrape for simple extraction tasks
- Use Agent when multiple steps or simulated human behavior are needed
- Switch to Headful for manual exploration, debugging, or direct interaction
- Leverage extraction scripts to process dynamic or shadow DOM content
- Parameterize with variables to make Agent tasks reusable across inputs
Understanding these task modes ensures you choose the most efficient and reliable workflow for your automation needs with Doppelganger.