domain-security-mcp-server

An MCP server that lets an AI agent audit the email and domain security of any domain — SPF, DKIM, DMARC, MTA-STS, TLS-RPT, BIMI, DNSSEC, DNS, TLS/SSL and WHOIS — in plain language. No API keys required.

Built on the v2 MCP SDK: the server speaks the 2026-07-28 protocol revision and keeps accepting 2025-era clients (Claude Desktop, Claude Code, Cursor) from the same factory — the era is negotiated per connection, so there is nothing to configure on either side.

Ask Claude "Is acme.com protected against email spoofing?" and it runs a full authentication audit and hands you a graded report with prioritised fixes — instead of you pasting a domain into five different web tools.

> Is ortamarco.me protected against email spoofing?

  email_auth_audit(domain="ortamarco.me")

  Grade: A (95/100) · MX: present
  ✅ SPF ends in '-all' (hard fail). 3/10 DNS lookups.
  ✅ DMARC policy is enforced ('p=reject').
  ✅ DKIM key found for selector: google.
  Top recommendation: add a TLS-RPT record for delivery-failure reports.

Why this exists

The email-security ecosystem is full of single-purpose web checkers (SPF here, DMARC there, WHOIS somewhere else) and the few MCP equivalents are locked behind paid API tokens. This server brings the whole deliverability & domain-security toolkit to any MCP client, key-free, with one headline workflow tool that does the synthesis for you.

It is the agent-facing companion to the network tools at ortamarco.me and shares the same battle-tested core (public-resolver DNS, host validation, timeouts).

Tools

| Tool | What it does | |---|---| | email_auth_audit ⭐ | One-call SPF + DKIM + DMARC + MX audit → 0–100 score, A–F grade, prioritised fixes | | spf_check | Parse SPF; recursively count DNS lookups vs the RFC 7208 limit of 10; flag +all/?all | | dmarc_check | Parse DMARC policy (p, sp, rua, pct, aspf/adkim) with warnings | | dkim_check | Probe <selector>._domainkey keys (supply selectors or use common ones) | | mta_sts_check | Validate the _mta-sts TXT and the .well-known/mta-sts.txt policy + mode | | tls_rpt_check | Check the _smtp._tls TLS-RPT record | | bimi_check | Check the default._bimi BIMI record | | dnssec_check | DS/DNSKEY presence + DNSSEC AD validation flag (via DoH) | | dns_lookup | All record types (A/AAAA/CNAME/MX/NS/TXT/SOA) via public resolvers | | ssl_certificate | TLS cert issuer, validity window, days-to-expiry, SANs, fingerprint | | whois_lookup | Registrar, dates, name servers, status (raw port-43 WHOIS, IANA-resolved) | | reverse_dns | PTR records for an IP | | ip_geolocation | Offline IP geolocation (DB-IP Lite) + reverse DNS | | mx_lookup | Mail servers (MX) with priority and resolved IPs | | caa_check | Which CAs may issue TLS certificates (CAA records) | | blacklist_check | IP/domain against open-access email DNSBLs | | dns_propagation | Compare a record across 5 public resolvers worldwide | | http_security_headers | Grade a site's HSTS, CSP, X-Content-Type-Options, X-Frame-Options, Referrer-Policy, Permissions-Policy and COOP | | analyze_email_headers | Parse raw headers → SPF/DKIM/DMARC verdicts + Received hop chain with delays |

Every tool is read-only, declares an outputSchema and returns structuredContent (validated by the SDK) alongside human-readable Markdown (default) or JSON (response_format="json"), plus actionable error messages.

Install

Requires Node.js 20.18+. Nothing to clone — every MCP client can run it with npx.

Use it with Claude Code

claude mcp add domain-security -- npx -y domain-security-mcp-server

Use it with Claude Desktop or Cursor

Add to claude_desktop_config.json (or ~/.cursor/mcp.json) — see examples/:

{
  "mcpServers": {
    "domain-security": {
      "command": "npx",
      "args": ["-y", "domain-security-mcp-server"]
    }
  }
}

On Windows use "command": "cmd" with "args": ["/c", "npx", "-y", "domain-security-mcp-server"]. Restart the client, then ask: "Audit the email security of stripe.com."

Self-host (HTTP transport)

The same server speaks stateless Streamable HTTP for remote or multi-client use. One endpoint serves both protocol eras and there is no session state, so no Mcp-Session-Id header is issued or expected.

TRANSPORT=http npx -y domain-security-mcp-server
# POST JSON-RPC to http://127.0.0.1:3000/mcp   ·   health at /healthz

It is safe by default: it binds to 127.0.0.1 and only accepts localhost Host and Origin headers, which blocks DNS-rebinding attacks from a web page. To expose it — for example behind Coolify or Traefik — opt in explicitly:

| Variable | Default | Purpose | |---|---|---| | TRANSPORT | stdio | http to serve Streamable HTTP | | PORT | 3000 | Listening port | | HOST | 127.0.0.1 | Bind address; 0.0.0.0 to accept remote connections | | ALLOWED_HOSTS | — | Comma-separated hostnames the Host header may carry (e.g. mcp.example.com) | | ALLOWED_ORIGINS | — | Comma-separated origins allowed to call from a browser | | MCP_AUTH_TOKEN | — | If set, every request needs Authorization: Bearer <token> |

Binding to a non-loopback address without ALLOWED_HOSTS or MCP_AUTH_TOKEN works, but the server says so on stderr. With Docker (the image sets HOST=0.0.0.0):

docker build -t domain-security-mcp .
docker run -p 3000:3000 -e ALLOWED_HOSTS=mcp.example.com -e MCP_AUTH_TOKEN=change-me domain-security-mcp

Security

The tools reach out to hosts that the caller names, so every outbound connection is screened against server-side request forgery:

• Private, loopback, link-local (cloud metadata), shared, multicast and reserved addresses are refused in every spelling, including IPv4 embedded in IPv6 ([::ffff:169.254.169.254]).
• The check happens at connect time, on the address the socket is actually about to use, so DNS rebinding and names only an internal resolver knows are refused too. Redirects are followed by hand and every hop is re-checked.
• Response bodies, redirects, WHOIS referrals and every network call are capped and time-limited.

Found a problem? Please open a private security advisory.

Develop

npm run dev      # tsx watch (stdio)
npm run inspect  # open the MCP Inspector against the built server
npm run build     # type-check + emit dist/
npm run typecheck # type-check only
npm test          # offline unit tests: SSRF guard, SPF/DMARC/DKIM scoring,
                  # header parsing, HTTP transport defaults
npm run smoke     # call all 19 tools on BOTH protocol eras and validate
                  # structuredContent against each tool's outputSchema

evals/ holds a 10-question LLM evaluation set (stable, verifiable) and instructions for running it — see evals/README.md.

How it works

src/
├── index.ts        # transport selection (stdio | http), v2 SDK entry points
├── server.ts       # factory: registers every tool on one McpServer
├── core/           # pure logic, no MCP coupling — reusable & testable
│   ├── validate.ts # input validation and the address classifier
│   ├── netguard.ts # connect-time SSRF guard, redirect-safe fetch, capped bodies
│   ├── dns.ts      # public-resolver DNS + DoH client
│   ├── net.ts      # MX, CAA, DNSBL and propagation checks
│   ├── tls.ts      # certificate inspection and trust
│   ├── whois.ts    # port-43 WHOIS with IANA/registrar referral
│   ├── http.ts     # security-header grading
│   ├── geoip.ts    # offline IP geolocation on DB-IP Lite (each file read on first use)
│   ├── email-headers.ts  # raw header parsing and hop timing
│   └── email-auth.ts  # SPF/DKIM/DMARC/MTA-STS/TLS-RPT/BIMI/DNSSEC + scoring
└── tools/          # thin MCP wrappers (Zod schemas, descriptions, formatting)

The core/ layer is deliberately free of any MCP types, so the exact same logic powers both this server and the web tools on ortamarco.me.

Credits

IP Geolocation by DB-IP. ip_geolocation uses the free DB-IP "IP to City Lite" database, licensed under CC BY 4.0 and installed as the @ip-location-db/dbip-city-mmdb package; the credit also appears in the tool's description and Markdown output. The database has no time zone, so time_zone is estimated from the coordinates with @photostructure/tz-lookup (CC0).

License

MIT © Marco Orta