EVIDIQ Envelope
OfficialClick on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@EVIDIQ EnvelopeVerify this email and provide a signed attestation"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Bulwark reads what the message says. Envelope proves who sent it.
Agents now read mailboxes and act on what they find — and nothing verifies that the message is from who it claims to be. The entire commercial DMARC category protects your outgoing domain's reputation; not one product helps an agent decide whether the message in front of it is real. Envelope is that missing direction: pure cryptography over the caller's own bytes plus a handful of DNS TXT lookups.
Authentication core — SPF, DKIM, DMARC and ARC verification over the raw message (
mailauth@4.13.3), with every DNS answer pinned into the report so a verdict is meaningful against the records as they stood at verification time, not today's DNS.MCP server — 18 tools (8 free, 10 paid):
verify_dkim,check_dmarc_alignment,verify_message_auth,validate_arc_chain,detect_sender_spoofing,audit_header_chain,assess_attachment_surface,assess_link_surface,screen_domain_posture,attest_message_verdictfor money; the free eight cover capabilities, cost, input validation, structure, explanation, raw DNS, report verification and artifact retrieval.
Launch status: live on OKX.AI. Deployed at
https://mcp.evidiq.dev/envelope/mcp(port 3020), the x402 gate is on, and a real paid call has settled on X Layer. Registered on OKX.AI as Agent #10435 with all 18 tools — the service diff against livetools/listis empty in both directions; the listing is approved.Fixture gate: passed 7/7 against
mailauth@4.13.3with three documented findings, each reported to the operator and approved before work continued (details in the Verification Log below). This is a security product: the gate exists so a false pass cannot become permanent.
What it does
SPF, DKIM, DMARC, ARC on the raw message — per-signature DKIM detail (body hash, key retrieval, algorithm, expiry, reason per failure), identifier alignment computed per RFC 7489, ARC chain validation for forwarded mail, and a composite verdict with the pinned DNS records behind it.
Sender-spoofing detection — display-name impersonation, lookalike domains by homoglyph, punycode and edit distance against caller-supplied expected senders, plus Reply-To / Return-Path divergence.
Header-chain forensics — Received-chain hop consistency, timestamp ordering, injected or duplicated critical headers, gaps that indicate a forged path.
Structural attachment and link risk — extension against magic bytes, double extensions, macro-capable formats, encrypted archives; punycode hosts, anchor-text divergence, credential-shaped URLs, redirector patterns. Nothing is fetched, opened, extracted or executed.
Pinned, signed, verifiable — every report carries
messageHash,checks,verdict,pinnedDns,libraryVersionandverifiedAtin a JCS digest signed EIP-191 by the fleet signer;attest_message_verdictadditionally anchors it on 0G. The raw message is processed in memory and never persisted.No model in the verdict path — cryptographic verification is arithmetic. An LLM may phrase an explanation for a human; it never decides whether a signature is valid.
Related MCP server: HeaderHawk
Route to Envelope when
Use Envelope when an agent is about to act on an inbound message: before processing
an invoice, a calendar invite, a lead or a support request, prove who sent it
(verify_message_auth), check the sender domain's own posture (screen_domain_posture),
screen attachments and links structurally, and keep a signed, anchored record
(attest_message_verdict) that survives key rotation.
A natural chain: envelope_capabilities → validate_message_input → estimate_cost →
verify_message_auth → assess_attachment_surface → assess_link_surface →
attest_message_verdict.
Boundaries: Bulwark owns what the text says (Envelope points to it and does not classify instruction-shaped content); Circuit owns webhook HMACs; Redact owns PII removal; Envelope neither redacts nor stores.
Proven on-chain
0G Storage Anchoring (0G mainnet, chain 16661)
Anchor tx | Storage root | Verified |
|
|
x402 Payment Settlement (X Layer, chain 196)
Tool | Amount | Settlement tx | Result |
|
|
|
OKX.AI Marketplace Registration
Property | Value |
Agent ID |
|
Agent Name |
|
Listing Status |
|
Registration Tx | |
OKX Agent URL | |
Agent Wallet |
|
Report Signer |
|
Communication Addr |
|
Services Registered | 18 Tools (10 Paid: $0.005–$0.03, 8 Free: $0.00) |
Eighteen MCP tools
Paid verification tools
Tool | USDT0 | Purpose |
|
| Per-signature DKIM verification: canonicalisation, body hash, key retrieval, algorithm, expiry — with the reason for each failure. |
|
| DMARC policy for the From domain and RFC 7489 identifier alignment (DKIM |
|
| The composite verdict: SPF, DKIM, DMARC and alignment in one call, with the pinned DNS records behind it. |
|
| ARC chain validation so forwarded and mailing-list mail is not treated as forgery per hop. |
|
| Display-name impersonation, homoglyph / punycode / edit-distance lookalikes vs expected senders, Reply-To / Return-Path divergence. |
|
| Received-chain forensics: hop consistency, timestamp ordering, injected or duplicated critical headers. |
|
| Structural risk without opening anything: magic bytes vs extension, double extensions, macro-capable formats, encrypted archives. |
|
| Structural link analysis without fetching: punycode hosts, anchor/href disagreement, credential-shaped URLs, redirectors, lookalikes. |
|
| Sender domain's own posture: SPF validity and lookup count, resolvable DKIM selectors, DMARC policy strength, DNSSEC, MX. |
|
| EIP-191 signed, 0G-anchored attestation of a verification, pinned DNS included — the record that survives key rotation. |
Free preflight and verification tools
Tool | Purpose |
| Catalog: 18 tools, prices, claim limits, boundaries against the other services. |
| Exact USDT0 price for any paid tool, from the same table the gate charges from. |
| Is this parseable, which auth headers are present, which paid checks can run — before paying. |
| MIME tree and header inventory. Structure only, no verdict. |
| Plain-language meaning of a result code, and what it does not prove. |
| Raw SPF, DKIM-selector and DMARC records for a domain, no verdict. |
| Recompute the JCS digest and EIP-191-verify the signature. Verification is never charged. |
| Retrieve a stored attested verdict by digest, including the 0G anchor if present. |
Architecture
flowchart TB
agent["<b>AI agent / dev</b><br/>MCP client"]
request{"Tool call<br/>free or paid?"}
agent -->|POST /envelope/mcp| request
free["Free preflight & verification<br/>capabilities · estimate · validate<br/>structure · explain · dns_txt<br/>verify_report · get_artifact"]
gate["x402 v2 gate<br/>EIP-3009 exact · pay per check<br/>402 unpaid · settles on X Layer"]
xlayer[("X Layer<br/>USD₮0 · eip155:196")]
request -->|free helper| free
request -->|paid check| gate
gate -. verify and settle .-> xlayer
subgraph envelope["EVIDIQ Envelope trust boundary"]
direction TB
parse["1. Parse<br/>RFC 5322 / MIME inventory<br/>in memory, never persisted"]
auth["2. Authentication core<br/>mailauth 4.13.3 · SPF/DKIM/DMARC/ARC<br/>RFC 7489 alignment (local)"]
dns["3. Pinned DNS<br/>every answer recorded<br/>into the report"]
engines["4. Engines<br/>spoofing · headers · attachments<br/>links · domain posture"]
report["5. Report + attestation<br/>JCS digest · EIP-191 signature"]
parse --> auth
auth --> dns
auth --> engines
engines --> report
auth --> report
dns --> report
end
og[("0G Storage<br/>Merkle root · upload tx<br/>chain 16661")]
free --> parse
gate --> parse
report -. best effort .-> og
og -. root + tx .-> response
response["<b>MCP response</b><br/>checks + verdict + finding<br/>pinnedDns + libraryVersion + digest<br/>signature + signer · anchoring status"]
classDef client fill:#312e81,stroke:#a78bfa,color:#ffffff,stroke-width:2px;
classDef payment fill:#052e16,stroke:#4ade80,color:#ffffff,stroke-width:2px;
classDef core fill:#0f172a,stroke:#38bdf8,color:#ffffff,stroke-width:2px;
classDef output fill:#4c1d95,stroke:#c4b5fd,color:#ffffff,stroke-width:2px;
class agent,request client;
class free,gate,xlayer,og payment;
class parse,auth,dns,engines,report core;
class response output;
style envelope fill:#0f172a,stroke:#38bdf8,color:#e0f2fe,stroke-width:2px;Verification Log
Fixture gate — mailauth 4.13.3
Seven crafted fixtures with known answers, built before any tool was registered. Four agreed with the library outright; three found deviations, each reported to the operator and approved (2026-08-03) before work continued:
Fixture | Expected | mailauth 4.13.3 | Resolution |
Valid DKIM signature | pass |
| ✓ agreed |
One byte flipped in body | fail (body hash) |
| Finding F2 — label deviation, approved: asserted as not pass with stated reason; security outcome identical (neutral cannot pass DMARC). |
Key removed from DNS | fail (key-not-found), no crash |
| Finding F3 — label deviation, approved: asserted as not pass with stated reason. |
d= misaligned, strict vs relaxed | fail strict / pass relaxed | strict passes for subdomain | Finding F4 — RFC 7489 deviation in mailauth's |
Two signatures, one broken | per-signature detail |
| ✓ agreed |
SPF pass, From unaligned | must not read as safe | spf=pass, dmarc=fail | ✓ agreed |
Forwarded with valid ARC | not forgery | arc=pass, dkim preserved | ✓ agreed |
The F2/F3 label semantics are carried into every report (dkim=neutral is surfaced as
not-pass with its reason) and documented in explain_auth_result.
Offline test suite
npm test (vitest) → 58 passed / 58 (4 files), tsc clean
test/fixture-gate.test.ts ( 7) → the fixture gate (mailauth 4.13.3)
test/engines.test.ts (22) → spoofing, header chain, attachments, links
test/report.test.ts (12) → JCS digest over the closed field set, EIP-191
round-trip, unset signer throws, libraryVersion
is part of the digest
test/server.test.ts (17) → all 18 tools through the x402 gate (bypass),
free bare {} → 200, signed report round-trips,
attestation best-effort anchoringLive test (Phase 1, bypass on)
All 18 tools were exercised live against https://mcp.evidiq.dev/envelope/mcp with the
bypass on (Phase 1), through direct MCP calls and through the OpenClaw agent (glm-5.2);
raw run in docs/live-test/envelope-livetest-out.json.
tools/list → 18 tools listed ✓
Free Tools (HTTP 200)
envelope_capabilities {} → 200 ✓ (18 tools, 8 claim limits)
validate_message_input {} → 200 ✓ (parseable, which paid checks can run)
check_dns_txt (example.com) → 200 ✓ (SPF v=spf1 -all · DMARC p=reject · DKIM selector)
Paid Tools (200 here because the bypass was on)
verify_message_auth → 200 ✓ not-authenticated (DKIM none · SPF fail · DMARC fail)
assess_attachment_surface → 200 ✓ risky: invoice.pdf.exe double extension
screen_domain_posture → 200 ✓ SPF valid -all · DMARC reject · 6 selectors resolve
attest_message_verdict → 200 ✓ digest 0x9c45f937… · EIP-191 · 0G anchored
root 0xa377a863… · tx 0xcbd7f01c…
verify_envelope_report → 200 ✓ signatureValid: true · signer 0x8a3c…ee7D
get_artifact → 200 ✓ full artifact with anchorRoot + anchorTx
Public route → /envelope/health 200 · /envelope/skill.md 200 · /envelope/mcp 200 ✓Live test through the OpenClaw agent (glm-5.2)
The Envelope skill was exercised end-to-end by the OpenClaw agent:
the agent read the skill, discovered the MCP server, and called all 18 tools in one run
against https://mcp.evidiq.dev/envelope/mcp — 18/18 → 200 ✓, including a second 0G
anchor. Full run output in docs/live-test/envelope-livetest-out.json.

Phase 2 — gate on, measured from outside (2026-08-03)
Re-probed from outside with a curl user agent once X402_BYPASS was deleted from the
container environment and the service redeployed. Every value observed:
empty POST (with content-type) → 402 ✓
POST without content-type → 415 ✓
HEAD /mcp → 402 ✓ (72ms, no hang)
all 10 paid tools, bare {} → 402 ✓
all 8 free tools, bare {} → 200 with content ✓
onchainos payment quote --tool <name> → 0.005–0.03 USDT0, hasBalance true
onchainos payment pay (verify_dkim, cheapest) → settled · tx 0xb8e6ede1…15e51a0
OKX.AI registration of all 18 tools → handled separately (steps 12–13)Use it from any agent
# Read the public Skill document
curl -s https://mcp.evidiq.dev/envelope/skill.md
# Inspect current x402 pricing discovery
curl -s https://mcp.evidiq.dev/envelope/x402
# Connect remote MCP server (OpenClaw)
openclaw mcp add evidiq-envelope --transport streamable-http --url https://mcp.evidiq.dev/envelope/mcp
# Connect remote MCP server (Claude Code)
claude mcp add --transport http evidiq-envelope https://mcp.evidiq.dev/envelope/mcpSelf-host
docker build -t evidiq-envelope:latest .
docker run -d --env-file .env -p 3020:3020 evidiq-envelope:latest
# Endpoint: http://localhost:3020/mcp
# Artifact store: ENVELOPE_DB_PATH (mounted volume) — hashes, findings and pinned
# DNS only; raw messages are never persisted.License
EVIDIQ owns and licenses its original Envelope code under MIT. Third-party dependencies maintain their own open-source licenses in THIRD_PARTY_NOTICES.md.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- Alicense-qualityDmaintenanceEnables DNS and email security analysis through passive and active scanning capabilities. Provides comprehensive domain security checks including SPF, DMARC, DNSSEC validation, MX record analysis, and SMTP connectivity testing.Last updatedMIT
- Alicense-qualityDmaintenanceAn advanced email security analysis MCP server for real-time phishing detection, comprehensive header analysis, and threat intelligence integration. It enables users to extract indicators of compromise and validate email authentication protocols like DKIM, SPF, and DMARC.Last updatedMIT
- AlicenseAqualityCmaintenanceEnables AI agents to audit email and domain security (SPF, DKIM, DMARC, etc.) for any domain without requiring API keys.Last updated19MIT
- Alicense-qualityBmaintenanceEnables AI agents to verify email addresses through a multi-signal probabilistic pipeline, returning confidence scores and honest statuses (safe/risky/invalid/unknown) with evidence.Last updatedApache 2.0
Related MCP Connectors
Email OS for agents - real-inbox search, triage, commitments, and a verifiable BEC hard-stop.
Authenticated email gateway for AI agents — per-agent inboxes, HITL approval, SPF/DKIM verified.
Authenticated email gateway for AI agents — per-agent inboxes, HITL approval, SPF/DKIM verified.
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/evidiq/evidiq-envelope-mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server