Render-fetch

★ 2

from sub-techie09

Drop-in upgrade to Claude's WebFetch that auto-detects JS-rendered pages and fetches them with Playwright headless browser

🔥🔥🔥🔥✓ VerifiedFreeQuick setup

#automation #devops #git

render-fetch

Drop-in upgrade to Claude Code's WebFetch — handles JS-rendered pages automatically.

Claude's built-in WebFetch doesn't execute JavaScript. Modern SPAs return empty skeletons. render-fetch auto-detects whether a page needs a real browser and routes accordingly — no config needed.

How it works

URL in
  │
  ├─ GET page (fast, reuses response)
  │
  ├─ Score for JS-rendering signals
  │     <div id="root/app/__next">  +3
  │     thin body (<500 chars)      +2
  │     data-reactroot / ng-version +2
  │     window.__REDUX_STATE__      +2
  │     bundled JS patterns         +1
  │     .md / docs domain           -3
  │
  ├─ Score < 3 → Static path
  │     HTTP → cheerio → Readability → markdown
  │
  └─ Score ≥ 3 → Browser path
        Playwright headless → networkidle → Readability → markdown

Tools

Tool	What it does	Key params	Returns
`fetch`	Fetch page as markdown (auto-detects mode)	`url`, `mode` (auto/static/browser), `maxLength` (default 50k), `timeout` (default 30s)	Clean markdown
`fetch_raw`	Fetch page as raw HTML	`url`, `mode`, `timeout`	Raw HTML string
`screenshot`	Take a viewport screenshot	`url`, `fullPage` (default false), `timeout`	Base64 PNG (max 2MB)

Why not always use the browser? Static path averages ~200ms. Browser path averages 2–3s plus Chromium launch overhead on first request.

Contributing

See CONTRIBUTING.md.

License

Installation

Add to your Claude Code settings (~/.claude/settings.json):

{
  "mcpServers": {
    "render-fetch": {
      "command": "npx",
      "args": ["-y", "render-fetch"]
    }
  }
}