Browser

Browser adds a command-line browser workflow to compatible coding agents. It helps an agent open pages, understand their current state, interact with controls, collect page content, and verify the result of each action. The same interface can work with a fresh local browser, an existing Chrome instance, or a hosted Browserbase session.

Highlights

• Reads structured page state so the agent can identify interactive elements without relying exclusively on images.
• Handles navigation, clicks, keyboard input, form fields, file uploads, scrolling, and multiple tabs.
• Captures screenshots and retrieves page text, markup, titles, URLs, and field values when needed.
• Offers local, existing-session, remote, and direct CDP connection options for different environments.
• Supports hosted sessions for workflows that need persistent identity, geographic routing, or stronger access capabilities.

This skill is useful for developers who want agents to inspect web applications, automate routine browser work, gather information, or reproduce interface problems. Install the Browserbase skills repository with the skills CLI; the browse package and a local Chrome-compatible browser are needed for local operation, while hosted sessions additionally require Browserbase credentials.