What problem does it solve?
agent-browser removes manual browser work by letting agents and scripts programmatically navigate websites, interact with forms and controls, capture visual evidence, and extract structured content so tasks like login automation, scraping, and UI testing can run reliably and repeatably.
Core Features & Use Cases
- Interactive snapshot refs (@e1, @e2, ...) that reduce context and enable precise element targeting for clicks, fills, and selections.
- Persistent named sessions, state save/load, and encrypted auth vault to reuse authenticated contexts for CI, monitoring, or long-running agents.
- Visual capture and diffing: full-page screenshots, annotated images, PDF export, video recording, and pixel diffs for visual regression testing.
- Robust wait strategies, CDP connection, proxy support, and profiling tools for debugging and performance analysis.
- Use cases include automated form submissions, authenticated scraping, end-to-end web testing, scheduled monitoring, and demo recording.
Quick Start
Open the page with agent-browser, take an interactive snapshot to capture refs, then perform fills, clicks, waits, and screenshots in sequence to complete the task.