TamingDNS MCP Server

Analyse a pasted block of raw email headers for spoofing and phishing tells. Input is the raw headers (everything up to the first blank line — "Show original" / "View source"); the message body is not required. Returns structured findings + a phishing verdict (likely_phishing/verify_sender/minor_concerns/legitimate/auth_incomplete) derived from SPF/DKIM/DMARC alignment (RFC 7489 §3.1), Microsoft composite-auth compauth reason codes (000/010 enforced failure, 601 intra-org implicit failure, 001 weak posture), ARC chain integrity (RFC 8617 §5.2), homoglyph / IDN look-alike From domains (Unicode UTS #39), header smuggling via duplicate singleton headers (RFC 5322 §3.6), display-name-as-address spoofing, hidden bidi/control characters (UAX #9), Received-chain timestamps running backward (RFC 5321 §4.4), link display-text-vs-href mismatches, originating-IP FCrDNS (RFC 8601), and RFC 8058 List-Unsubscribe deliverability posture (category:compliance, never moves the verdict). Findings include DMARC_FAILED, SPF_HARD_FAIL, DKIM_SIGNATURE_FAILED, COMPAUTH_EXPLICIT_FAILURE, COMPAUTH_INTRA_ORG_IMPLICIT_FAILURE, HOMOGLYPH_DOMAIN, DUPLICATE_HEADER, DISPLAY_NAME_SPOOFED_ADDRESS, RECEIVED_TIMESTAMPS_BACKWARD, URL_DISPLAY_MISMATCH, HIDDEN_CONTROL_CHAR, BRAND_IMPERSONATION, and more (registered in docs/finding-codes.md). `status` is the finding-severity aggregate; `verdict` is the authoritative phishing judgment; `analysis` carries the verdict object + per-mechanism authentication. For privacy the echoed `input` is a digest only (from / subject / message_id) — raw headers are never returned. Read-only.

Check default._bimi.<domain>: parses the BIMI record, validates the SVG Tiny-PS logo, inspects the VMC chain (expiry, EKU, SCT, issuer-accepted, SAN match, logotype hash), and verifies the DMARC prerequisite (p=quarantine/reject at pct=100 with no sp=none — the RFC 7489 reading Gmail and Apple Mail evaluate today). DMARCbis (RFC 9989) aware: t= is parsed into analysis.dmarc.t, t=y emits TESTING_MODE (warn under p=quarantine — bis receivers read it as none; info under p=reject), and a present pct= emits DEPRECATED_TAG (removed by RFC 9989 Appendix A.6, advisory only). Returns grade (A-F), structured findings, and an `analysis` block (logo/vmc/dmarc/provider_matrix). Read-only.

Inspect a domain's MX / mail-routing posture. Resolves the apex MX RRset (priority-sorted), and for each host the A/AAAA addresses, PTR + forward-confirmed reverse DNS (FCrDNS), DANE TLSA at _25._tcp.<host>, IP-literal (RFC 5321 §5) and CNAME (RFC 2181 §10.3) violations, and the mail provider. Detects the null MX (RFC 7505), a null MX wrongly mixed with real MX, and the implicit MX via the apex A/AAAA when no MX exists (RFC 5321 §5). Probes the MTA-STS and TLS-RPT prerequisites over DNS (case-sensitive per RFC 8461/8460) with outage reported as unknown, never absence. Returns grade (A-F) with grade_reasons, a verdict (healthy/warnings/misconfigured/null_mx/no_mx/implicit_mx), structured findings (NO_MX_RECORD, MX_IMPLICIT_FALLBACK, NULL_MX, NULL_MX_WITH_OTHER, MX_POINTS_TO_IP, MX_POINTS_TO_CNAME, MX_HOSTNAME_UNRESOLVABLE, NO_PTR_RECORD, FCRDNS_MISMATCH, MX_IPV6_ONLY, MX_NO_TRANSPORT_SECURITY, MX_GATEWAY_RDNS_COSMETIC, MX_SINGLE_HOST, MX_DUPLICATE_HOST, MX_SHARED_PRIORITY, MX_MTA_STS_WITHOUT_TLS_RPT), and an `analysis` block (per-host enrichment incl. provider, ips with FCrDNS status, dane_tlsa, and the prerequisite probe results). DNSSEC AD is a UI-only signal, not gradeable through the DoH contract. Read-only.

Check SMTP DANE / TLSA records for a domain (RFC 7672). Resolves the MX hosts (falls back to the apex when none), looks up the TLSA record at _25._tcp.<mx-host> over binary DoH, classifies each record (usage 0-3 / selector 0-1 / matching 0-2 per RFC 7671 §2.1), and confirms the answer carried the DNSSEC AD bit — DANE is meaningless without DNSSEC, so an unsigned TLSA answer is the dominant failure mode. Returns grade (A-F) with grade_reasons, a verdict (valid/valid_with_warnings/no_dnssec/not_configured/error), structured findings (DANE_NO_TLSA, DANE_TLSA_HOST_UNSIGNED, DANE_PKIX_TA_USED_WITHOUT_CHAIN, DANE_TLSA_EXACT_MATCH, DANE_PARTIAL_COVERAGE, DANE_TLSA_INVALID_USAGE/SELECTOR/MATCHING), and an `analysis` block (per-host TLSA records + DNSSEC status, the spec-compliance signals, and the MTA-STS cross-signal). DNS-only: it never opens a port-25 STARTTLS handshake, so it proves the TLSA records exist, are well-formed, and are DNSSEC-signed — never that they match the live cert (Hostinger and Cloudflare both block outbound 25). Read-only.

Look up a DKIM public key at <selector>._domainkey.<domain> and validate it. Detects key revocation (empty p=), RSA key length (KEY_TOO_SHORT below 1024, KEY_WEAK below 2048), SHA-1 hash deprecation, and t=y test mode. For selector1/selector2 it also checks a Microsoft 365 CNAME target carries the token Microsoft builds for the domain (M365_DKIM_CNAME_MISMATCH), even when no key resolves. Returns an `analysis` block with the per-key facts (key_type, key_bits, key_strength, revoked, test_mode, hash_algs, m365_cname). Read-only.

Check the _dmarc record for a domain with DUAL discovery: the RFC 7489 §6.6.3 organisational-domain ladder (what major receivers enforce today — grade/verdict anchor to it) AND the RFC 9989 §4.10 DNS Tree Walk (8-query cap, psd=n/y stop rules, fewest-labels org selection) into analysis.discovery {legacy, tree_walk, diverges}. When the two methods select different policy records or org domains, DISCOVERY_DIVERGENCE (warn) fires with a one-notch grade downgrade. Also DMARCbis tag-aware: t=/np=/psd= parsed into analysis.tags, removed tags (pct/rf/ri) emit DEPRECATED_TAG, t=y emits TESTING_MODE with a one-notch downgrade. Returns the grade (A-F) with grade_reasons, a verdict (enforced/partial/monitoring/monitoring_blind/invalid/not_configured), structured findings (NO_DMARC_RECORD, MULTIPLE_DMARC_RECORDS, INVALID_TAG, POLICY_NONE, POLICY_NONE_NO_RUA, PCT_BELOW_100, RUA_MISSING_UNDER_ENFORCEMENT, SP_NOT_ENFORCED, EXTERNAL_RUA_UNAUTHORIZED, DMARC_INHERITED_FROM_ORG, TESTING_MODE, DEPRECATED_TAG, DISCOVERY_DIVERGENCE), and an `analysis` block with the lookup ladder, discovery, source (self/org), org_domain, normalised tags, per-recipient RUA/RUF external-authorisation detail (RFC 7489 §7.1), and rua_unauthorised. Read-only.

Validate a domain's DNSSEC chain of trust. Pulls DS (at the parent), DNSKEY, and RRSIG over binary DoH, walks the chain zone-by-zone (root → TLD → domain) recomputing each DS digest against the published DNSKEYs (RFC 4034 §5.1.4), and cross-checks the AD bit across Cloudflare, Google, Quad9, and OpenDNS so a single resolver's caching can't mask a break. Returns grade (A-F) with grade_reasons, a verdict (signed/signed_no_ds/partial/unsigned/broken), structured findings (DNSSEC_UNSIGNED, DNSSEC_DS_MISSING_AT_PARENT, DNSSEC_DS_ALGORITHM_DEPRECATED, DNSSEC_DNSKEY_ALGORITHM_DEPRECATED, DNSSEC_RRSIG_EXPIRED, DNSSEC_BOGUS, DNSSEC_AD_BIT_MISSING), and an `analysis` block (the full chain-walk, the per-resolver validator matrix, NSEC3 iteration/opt-out, CDS/CDNSKEY publication, signing mode, root-anchor state). Unsigned is the common case and returns status "info" (DNSSEC is optional), not a failure; signed_no_ds (DNSKEY published, no DS at the parent yet) is treated by validators as unsigned and also returns "info"; a published-but-non-validating zone returns "fail". Read-only.

Check MTA-STS (RFC 8461) for a domain: the _mta-sts TXT record (strict §3.1 selection — case-sensitive leading v=STSv1, exactly one record, required 1-32 alphanumeric id), the policy file at https://mta-sts.<domain>/.well-known/mta-sts.txt fetched WITHOUT following redirects (§3.3), strict §3.2 policy validation (version/mode/max_age required, mx required unless mode none, max_age ≤ 31557600, duplicate keys first-wins), the policy-host X.509 certificate (§3.3 — an invalid cert means compliant senders fetch no policy), and live MX coverage against the mx: patterns (§4.1 — wildcards match exactly one leftmost label; coverage is all/partial/none, partial meaning strict senders skip the uncovered hosts per §5.1). Returns grade (A-F) with grade_reasons, a verdict (enforced/testing/reporting_only/invalid/not_configured), structured findings (MTA_STS_NO_TXT_RECORD, MTA_STS_RECORD_INVALID, MULTIPLE_MTA_STS_RECORDS, MTA_STS_POLICY_UNREACHABLE, MTA_STS_POLICY_REDIRECT, MTA_STS_POLICY_INVALID, MTA_STS_CERT_INVALID, MTA_STS_MODE_NONE, MTA_STS_MODE_TESTING, MTA_STS_MX_MISMATCH, MTA_STS_MX_PARTIAL_COVERAGE, MTA_STS_MAX_AGE_TOO_LOW), and an `analysis` block (record, policy incl. fetch_stage, cert, MX coverage detail, TLS-RPT/DANE cross-signals). Read-only.

Check the SPF record for a domain. Returns the raw record, grade (A-F), the terminal `all` qualifier (resolved through redirect= chains), structured findings (NEAR_LOOKUP_LIMIT, LOOKUP_LIMIT_EXCEEDED, PLUS_ALL_DANGEROUS, PTR_MECHANISM_DEPRECATED, etc.), and an `analysis` block with the expanded include/redirect tree and the RFC 7208 §4.6.4 lookup count. Read-only.

Check the TLS-RPT (SMTP TLS Reporting, RFC 8460) record at _smtp._tls.<domain>. Strict §3 classification: the record must BEGIN with the case-sensitive v=TLSRPTv1 tag, more than one such record means receivers treat TLS-RPT as not implemented, and rua= is required. Validates every rua= destination syntactically (mailto:/https:// — reachability is never probed), and detects the MTA-STS and DANE prerequisites over DNS (TLSA at _25._tcp of the lowest-preference MX) with outage reported as unknown, never as absence. Returns grade (A-F) with grade_reasons, a verdict (valid/partial/invalid/not_configured), structured findings (TLS_RPT_NO_RECORD, TLS_RPT_INVALID_VERSION, TLS_RPT_MULTIPLE_RECORDS, TLS_RPT_NO_RUA, TLS_RPT_RUA_INVALID_URI, TLS_RPT_NO_VALID_RUA, TLS_RPT_NO_ENFORCEMENT_POLICY), and an `analysis` block with parsed tags, per-URI rua validation detail, and the prerequisite probe results. Read-only.