Connect your agent to every tool on this site

The endpoint

Streamable HTTP, the current MCP transport — one URL, no separate SSE endpoint to configure:

https://tools.scult.in/api/mcp

Connect

Claude Code

One command, from any terminal:

claude mcp add --transport http scult-tools https://tools.scult.in/api/mcp

Claude Desktop, Cursor, and other JSON-config clients

Add this to the client's MCP config file (Claude Desktop's claude_desktop_config.json, or Cursor's .cursor/mcp.json):

{
  "mcpServers": {
    "scult-tools": {
      "url": "https://tools.scult.in/api/mcp"
    }
  }
}

If a client only supports stdio servers and can't dial Streamable HTTP directly, bridge it with mcp-remote instead:

{
  "mcpServers": {
    "scult-tools": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://tools.scult.in/api/mcp"]
    }
  }
}

ChatGPT

Add a custom connector from ChatGPT's settings and paste the endpoint URL above — no config file involved.

What's callable — 25 tools

Pure-compute tools

Same logic as the tool pages themselves — no network call, generous rate limit.

• generate_schema_markupJSON-LD for 9 schema.org types
• generate_faq_schemaFAQPage JSON-LD + visible HTML block
• build_utm_urlGA4-safe UTM campaign URLs
• calculate_marketing_roiCampaign ROI and ROAS side by side
• compute_invoice_totalsInvoice subtotal, discount, tax, total
• generate_business_namesDeterministic brand-name candidates
• generate_slogansTaglines by tone, with ad-fit checks
• generate_email_signatureEmail-safe HTML signature, 3 layouts
• format_jsonFormat, minify or repair JSON
• count_wordsWord/character/reading-time + keyword density
• generate_color_paletteOKLCH harmony + WCAG contrast, exportable
• generate_qr_codeQR PNG for a URL, text, WiFi or UPI payment
• generate_faviconFull favicon set from a logo image

Server-backed tools

These fetch a real target site or call Google's PageSpeed Insights — same implementation as AI Visibility Checker and Website Speed Test, with a stricter, separate rate limit.

• check_ai_visibilityA site's AI-crawler/answer-engine visibility, scored
• test_website_speedReal PageSpeed Insights / Lighthouse run

Prompt Library

• search_promptsKeyword search across 1,170+ prompts, template body included
• get_promptFull prompt: template, variables, why it works
• list_prompt_categoriesAll prompt categories and groups

Skills Library

• search_skillsKeyword search across the Skills Library
• get_skillFull skill body — withheld where the source license is unconfirmed
• list_skill_categoriesAll skill categories with live counts

Guides & Blog

• list_guidesThe evergreen how-to guides — a small set, separate from the blog
• get_guideOne guide in full, every section
• search_blogKeyword search across 200+ long-form posts, full body included
• get_blog_postOne post in full: sections, FAQ, sources, related links

Rate limits

Two limits, tracked per connection, independent of the browser rate limits the website itself uses:

• 30 calls / 60 seconds for everything except the two server-backed tools — generous for an agent making several calls in one turn, tight enough to stop a runaway loop.
• 3 calls / 60 seconds for check_ai_visibility and test_website_speed — stricter than the website's own limit, since each one spends real third-party quota.

Hitting a limit returns a normal tool result stating how many seconds until the next call is allowed — never a silent failure.

Worth knowing

• get_skill withholds a skill's body when its source license isn't confirmed to permit redistribution, returning a source link instead — the same rule the Skills Library's own pages enforce, not a separate one invented for this server.
• No authentication and no per-agent identity: this is a public, rate-limited server, matching this site's own no-signup posture. Don't send secrets through it.
• Everything here calls the exact same code the tool pages use — there is no separate "MCP version" of any tool's logic to drift out of sync.