Steel Puppeteer

Example Use Cases

• 1

  Navigate and extract structured or full-page content from live websites, including handling lazy-loading via automated scrolling.

• 2

  Perform interactive UI workflows (clicks, fills, selects, hovers) to complete tasks such as form submission, data entry, or multi-step navigation.

• 3

  Capture screenshots and browser console logs or execute arbitrary JavaScript in a real browser for debugging, validation, or dynamic content evaluation.

Description

Steel Puppeteer provides a Model Context Protocol server that integrates Puppeteer with Steel to enable real browser automation for LLM-driven agents. It supports navigation, clicking, form filling, selecting, hovering, scrolling for lazy-loaded content, JavaScript evaluation, content extraction with token-aware handling, console log capture, and screenshot storage. The server can operate with a local Steel instance or the Steel cloud and is designed to let agents interact with and observe real webpages for richer, grounded actions and diagnostics.