Steel Puppeteer
@MCP-Mirror
Mirror of
Overview
What is Steel Puppeteer?
Steel Puppeteer is a Model Context Protocol server that provides browser automation capabilities using Puppeteer and Steel. It enables LLMs to interact with web pages, take screenshots, and execute JavaScript in a real browser environment.
How to use Steel Puppeteer?
Install dependencies with npm install, build with npm run build, and start the server with npm start. Configure it for Claude Desktop by adding a steel-puppeteer entry to claude_desktop_config.json with the compiled index.js path and appropriate environment variables (STEEL_LOCAL, STEEL_API_KEY, STEEL_URL).
Key features of Steel Puppeteer
- Browser automation with Puppeteer
- Steel integration for session management
- Console log monitoring and capture
- Screenshot capture of pages or elements
- JavaScript execution in the browser
- Web interaction: navigation, clicking, form filling
- Content extraction with token limit handling
- Lazy-loading support through scrolling
Use cases of Steel Puppeteer
- Have an LLM navigate to a URL and take a screenshot.
- Automate form filling and clicking on a web page.
- Extract page content or execute custom JavaScript for data mining.
- Monitor browser console logs during automated testing.
- Scroll through lazy-loaded pages to capture all content.
FAQ from Steel Puppeteer
What environment variables are required?
STEEL_LOCAL (optional, default "false"): set to "true" for local Steel instance. STEEL_API_KEY required only if STEEL_LOCAL is "false". STEEL_URL optional for custom Steel deployment.
How do I run Steel Puppeteer locally?
Install dependencies (npm install), build (npm run build), then start (npm start). If using a local Steel instance, set STEEL_LOCAL=true in a .env file.
What tools does Steel Puppeteer provide?
It provides nine tools: puppeteer_navigate, puppeteer_screenshot, puppeteer_click, puppeteer_fill, puppeteer_select, puppeteer_hover, puppeteer_evaluate, puppeteer_get_content, and puppeteer_scroll.
What resources are exposed?
Two resource types: console logs (console://logs) for browser console output, and screenshot images (screenshot://<name>) captured during sessions.
