What problem does it solve?
You need reliable, repeatable browser automation to open pages, perform multi-step actions, and extract structured information without manual clicking.
Core Features & Use Cases
- Command live browsing: Supports
navigate, act, extract, observe, screenshot, and close to complete common browsing workflows.
- Natural-language interaction: Uses Stagehand to interpret instructions for clicking, navigation flows, and interaction goals.
- Structured extraction: Optionally applies a JSON-to-Zod schema to return validated structured results.
- Visual evidence: Captures screenshots after actions and returns the saved screenshot path to the user.
- Real-world use case: Open a target website, click through to a specific item, extract key fields (e.g., title, date, and summary) into a structured response, and attach screenshots for review.
Quick Start
Ask the assistant: "Use the browser skill to navigate to https://news.ycombinator.com/ and screenshot the results."