ToolCenter MCP
The Swiss-Army MCP server for web tools — one install, 15 LLM-ready tools.
Give your AI agent instant access to web search, scraping, screenshots, SEO/accessibility audits, DNS/WHOIS/SSL checks, and more — all through a single Model Context Protocol server. Outputs are optimized for LLM consumption (clean Markdown, structured reports) so your agent spends tokens on reasoning, not parsing HTML soup.
Tools included
| Tool | What it does |
|---|---|
web_search | Search the web (aggregated engines) — news/images/science categories, time filters |
scrape_url | Fetch a page and return clean Markdown (boilerplate stripped via Readability) |
screenshot | Capture a page — full-page, device emulation, dark mode, ad-blocking |
get_metadata | Title, description, Open Graph, Twitter Card, canonical, favicon |
url_to_pdf | Render any URL to PDF — page size, orientation, headers/footers |
website_diff | Compare two URLs — added/removed/modified sections with similarity score |
analyze_seo | SEO audit with scored report, issues grouped by severity, fixes |
analyze_accessibility | axe-core a11y audit, WCAG violations grouped by impact |
check_broken_links | Scan a page for broken (4xx/5xx/timeout) links |
detect_tech_stack | Wappalyzer-style stack fingerprinting |
preview_link | Slack/Discord-style unfurl card for a URL |
dns_lookup | DNS records — A, AAAA, MX, NS, TXT, CNAME, SOA, SRV, CAA, PTR |
whois_lookup | Registrar, creation/expiry, name servers, ownership |
check_ssl | TLS cert — issuer, validity, days-until-expiry, SANs, cipher |
check_status | HTTP-ping — up/down, status, response time, redirects |
Install
1. Get an API key
Create a free account at toolcenter.dev and copy your API key from the dashboard.
2. Add to your MCP client
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_xxxxxxxxxxxxxxxxxxxx"
}
}
}
}
Restart Claude Desktop. The 15 tools appear under the hammer icon.
Cursor
Settings → MCP → Add server:
{
"toolcenter": {
"command": "npx",
"args": ["-y", "toolcenter-mcp"],
"env": { "TOOLCENTER_API_KEY": "tc_..." }
}
}
Claude Code (CLI)
claude mcp add toolcenter -e TOOLCENTER_API_KEY=tc_... -- npx -y toolcenter-mcp
3. (Optional) Self-host the backend
The MCP server talks to api.toolcenter.dev by default. To point at your own ToolCenter instance, set TOOLCENTER_BASE_URL:
"env": {
"TOOLCENTER_API_KEY": "tc_...",
"TOOLCENTER_BASE_URL": "https://api.your-domain.com"
}
Quickstart: build an agent in 5 minutes
Ask Claude:
Research the top 3 competitors of Linear.app. For each, give me their pricing page, tech stack, SEO score, and a screenshot of their homepage.
The agent will chain web_search → scrape_url → detect_tech_stack → analyze_seo → screenshot automatically. No glue code.
Development
git clone https://github.com/toolcenter-dev/mcp
cd mcp
npm install
cp .env.example .env # add your key
npm run build
TOOLCENTER_API_KEY=tc_... node dist/index.js # runs over stdio
Output design
Every tool returns Markdown structured for LLM consumption. For example, scrape_url runs the HTML through Mozilla Readability to strip nav/footer/ads, then converts the article body with Turndown. A 2 MB page becomes ~4 KB of Markdown — 500× fewer tokens than raw HTML.
License
MIT © 2026 ToolCenter