browser-use/workflow-use

What it solves

Workflow Use provides a way to create deterministic, reliable browser automation workflows that avoid the fragility of traditional RPA (Robotic Process Automation) and the unpredictability of pure LLM-based browser agents. It allows users to record a task once and reuse it indefinitely with different variables, ensuring the process is executed exactly as intended without needing to constant prompting.

How it works

The system converts browser interactions into structured, semantic workflows stored as JSON files. It offers three primary ways to create these workflows:

1. Recording: Users can record their own interactions via a browser extension.
2. Generation Mode: Users describe a task in natural language; the system runs browser-use to complete the task once, then converts that execution history into a reusable semantic workflow.
3. Programmatic Creation: Workflows can be created and managed via a Python API.

When executing, the system can run these workflows without AI (using semantic mapping) or use an LLM as a fallback mechanism to "heal" the workflow if a step fails.

Who it’s for

Developers and automation engineers who need reliable, repeatable browser-based tasks (like filling out forms or scraping specific data) that are more stable than standard scripts but easier to create than manual coding.

Highlights

• Record Once, Reuse Forever: Converts recordings into deterministic workflows that automatically extract variables from forms.
• Generation Mode: Automatically creates reusable workflows from a simple natural language description.
• Cloud Browser Support: Ability to run workflows in the Browser-Use Cloud with semantic abstraction to eliminate LLM calls during execution.
• Visual Management: Includes a GUI to visualize workflows as interactive graphs, monitor logs, and execute tasks with custom parameters.
• Self-Healing Foundation: Designed to fallback to AI agents to fix broken steps automatically when a website's layout changes.