MCP
+
TC

Integration Guide

MCP Server + ToolCenter

Give your AI agent 15 web tools in one install. Search, scrape, screenshot, audit, diff — all through a single Model Context Protocol server. Works with Claude Desktop, Cursor, Claude Code and any MCP client.

Get Free API Key View on GitHub
📦 npm: toolcenter-mcp · ⚖️ MIT · 🔧 15 tools

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI agents securely access tools and data sources through a unified interface. Instead of wiring five separate APIs every time you build an agent, you install one MCP server and get instant, typed access to its tools.

ToolCenter's MCP server exposes 15 curated web tools — the ones agents actually need — with outputs optimized for LLM consumption. A 2 MB HTML page becomes ~4 KB of clean Markdown. Your agent spends tokens on reasoning, not parsing.

Why ToolCenter MCP?

🧰

One Install, 15 Tools

No more stitching together 5 APIs. Search, scrape, screenshot, SEO audit, DNS, SSL — all behind a single config block.

📝

LLM-Optimized Outputs

Every tool pipes through an adapter: Mozilla Readability + Turndown for clean Markdown, grouped-by-severity reports for audits. 500× fewer tokens than raw HTML.

🔓

Open Source, Self-Hostable

MIT licensed on GitHub. Point at our managed API or your own ToolCenter instance via TOOLCENTER_BASE_URL.

Install in 60 seconds

Pick your MCP client below.

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "toolcenter": {
      "command": "npx",
      "args": ["-y", "toolcenter-mcp"],
      "env": {
        "TOOLCENTER_API_KEY": "tc_..."
      }
    }
  }
}

Cursor

Settings → MCP → Add server:

{
  "toolcenter": {
    "command": "npx",
    "args": ["-y", "toolcenter-mcp"],
    "env": {
      "TOOLCENTER_API_KEY": "tc_..."
    }
  }
}

Claude Code (CLI)

One-liner — adds the server to your user config:

claude mcp add toolcenter \
  -e TOOLCENTER_API_KEY=tc_... \
  -- npx -y toolcenter-mcp

Other MCP clients

Any stdio-speaking MCP client works — pass the env var and command:

TOOLCENTER_API_KEY=tc_... \
  npx -y toolcenter-mcp

The 15 tools

🔍 web_search

Search the web (aggregated engines, time filters)

📄 scrape_url

Fetch a page as clean Markdown (Readability + Turndown)

📸 screenshot

Capture any URL — full page, device emulation

🏷️ get_metadata

Title, description, Open Graph, Twitter Card

📕 url_to_pdf

Render URL to PDF — size, orientation, margins

🔀 website_diff

Compare two URLs — added/removed/modified

📈 analyze_seo

Scored SEO report — issues grouped by severity

♿ analyze_accessibility

WCAG audit — violations by impact

🔗 check_broken_links

Scan a page for 4xx/5xx/timeout links

🧬 detect_tech_stack

Wappalyzer-style framework fingerprinting

🎴 preview_link

Slack-style unfurl card for a URL

🌐 dns_lookup

A, AAAA, MX, NS, TXT, CNAME, SOA, SRV, CAA

📋 whois_lookup

Registrar, creation/expiry, ownership

🔒 check_ssl

TLS cert — issuer, validity, days to expiry

🟢 check_status

HTTP ping — status, response time, redirects

Build a competitor research agent in 5 minutes

Ask Claude (or any MCP-enabled client):

"Research Linear's top 3 competitors. For each, give me their pricing page content, tech stack, SEO score, and a screenshot of their homepage. Summarize in a table."

The agent automatically chains these MCP tools:

  1. 1. web_search — find the competitors
  2. 2. scrape_url ×3 — read each pricing page as Markdown
  3. 3. detect_tech_stack ×3 — fingerprint their stack
  4. 4. analyze_seo ×3 — score each site
  5. 5. screenshot ×3 — capture their homepages

No glue code. No SDK setup. No 5 separate API keys.

Start building smarter agents

Free tier includes 100 calls/month. Upgrade anytime.

Get Free API Key View on GitHub