VibeBrowser Browser Automation CLI

Problem

When an AI agent needs to inspect real pages, Playwright and Puppeteer can add dependency, version, and browser-driver overhead. vibe-browser connects to Chrome, Chromium, Brave, or Edge via CDP, turning page reads, interactions, and screenshots into a lighter CLI workflow for research, development checks, test data collection, and small-scale page sampling.

Core Capabilities

• Read pages: use get text, get html, and selectors to inspect structure; get html is capped around 50 KB by default to avoid large SPAs overflowing context.
• Interact and wait: support clicks, typing, and form handling, with wait selector and wait url for element or route changes.
• Browser discovery: resolve local browser paths and user-data directories, helping connect to existing profiles.
• Diagnostic output: when no interactive elements exist, it reports top-level node count, guiding a narrower selector or non-interactive retry.

Boundaries

This is a CLI-style browser control tool, not a full crawler; be careful with strong anti-bot controls, captchas, and persistent login state. On Windows, long-running daemon mode is unavailable, so use --cdp-url or launch the browser per command.