Description
Browser Agent is an AI-powered browser automation tool that accepts natural language goals and executes them against a live Chromium session. Each turn of an agent run produces a screenshot, DOM snapshot, and reasoning trace as verifiable evidence of the actions taken. The tool supports E2E test runs, visual regression comparison against prior baselines, design audits with design token extraction, and parallel batch execution with self-healing recovery for stuck navigation states. It is distributed as an MIT-licensed npm package with a CLI for one-off runs and a TypeScript SDK for CI pipeline integration.