High-speed MCP tool for Claude context optimization. Strips tracking scripts, footers, and HTML noise, returning clean Markdown to save up to 70% in token overhead. Powered by Bintu.

What you get:

Article/main content preferred; nav, header, footer, scripts stripped by default
Headings, tables, and links preserved as Markdown -6,000-character text cap with a truncated flag
Billing, rate limits, and analytics handled entirely on this page

Quick start

Click Subscribe and pick a tier below.
Copy your MCPize API key (sk_...). No upstream Bintu key needed — the gateway injects your credential automatically.
Connect your client:

Claude Code:

claude mcp add --transport http bintu https://bintu-web-reader.mcpize.run --header "Authorization: Bearer YOUR_MCPIZE_KEY"

Claude web/Desktop: add a custom connector with URL https://bintu-web-reader.mcpize.run and choose "Sign in now" — you'll sign in with your MCPize account and approve access once. (If your plan shows a request-header option, the sk_… key works there too.)

Cursor / other MCP clients: connect to https://bintu-web-reader.mcpize.run with header Authorization: Bearer YOUR_MCPIZE_KEY.

Call the tool:

Tool: bintu_scrape_text Arguments: {"url": "https://www.iana.org/domains/reserved", "render_js": false}

Expect:

{"content": [{"type": "text", "text": "# IANA-managed Reserved Domains\n## Example domains\n..."}], "isError": false}

If isError is true, read the text — it tells the model how to self-correct (e.g. bad URL, rate limit, render_js_disabled → retry with false).

Limits: public HTML pages only · 6,000-character text cap · plain-HTML extraction (render_js currently disabled) · no logins, paywalls, or CAPTCHAs.

Full docs and examples: https://github.com/l20941485-blip/Bintu-API

LLM Web Scraper & Clean Text Extractor

How to pay

Subscribe