Skip to main content

​
OpenSteer

Browser automation framework for developers and AI agents with deterministic replay. OpenSteer gives you one API for local and cloud runs, description-based actions, structured extraction, and CUA agent workflows.

Installation

Get started with OpenSteer in minutes

Quick Start

Run your first automation script

API Reference

Complete SDK and CLI documentation

Examples

Real-world automation examples

​
What It Does

OpenSteer provides a unified browser automation framework that works seamlessly across local and cloud environments. Whether you’re building automated tests, scraping data, or creating AI-powered browser agents, OpenSteer gives you the tools you need. Unified Local/Cloud Execution - Use the same API surface for both local and cloud runs. Switch between environments without changing your code. Descriptor-Aware Actions - Use natural language descriptions to interact with web pages. OpenSteer automatically persists selectors for deterministic replay. Structured Extraction - Extract data from web pages using typed schemas. Get structured, type-safe results every time. CUA Agent Support - Built-in support for Computer Use Agents with OpenAI, Anthropic, and Google models.

​
Key Features

​
Robust Selectors with Persistence

OpenSteer’s descriptor-aware selectors use LLM-powered resolution with automatic caching. When you describe an element once, the selector is persisted in .opensteer/selectors/<namespace> for reliable replay. 
await opensteer.click({ description: "main call to action" });

​
Unified API Surface

One API works everywhere - local development, CI/CD, and cloud execution: 
const opensteer = new Opensteer({ 
  name: "my-script",
  cloud: true // Toggle cloud mode
});

​
HTML Optimization

Automatic HTML cleaning and snapshot modes optimize for different use cases:
  • action mode: Optimized for element interaction
  • extraction mode: Optimized for data extraction
  • Removes noise while preserving semantic structure

​
CLI and SDK Interfaces

Use OpenSteer as a CLI tool for interactive sessions or as an SDK for programmatic automation: 
# CLI mode
opensteer open https://example.com --session demo
opensteer click --description "login button" --session demo

// SDK mode
const opensteer = new Opensteer({ name: "demo" });
await opensteer.launch();
await opensteer.goto("https://example.com");
await opensteer.click({ description: "login button" });

​
Use Cases

  • Automated Testing: Create maintainable browser tests with natural language selectors
  • Data Extraction: Extract structured data from web pages at scale
  • AI Agents: Build browser automation agents that can adapt to changing UIs
  • Workflow Automation: Automate repetitive browser tasks with deterministic replay
  • Quality Assurance: Validate web applications with description-based assertions

​
Next Steps

Install OpenSteer

Set up your environment and install dependencies

Quick Start Guide

Write your first automation script in 5 minutes