Mark Reynolds
Post-quantum document anchoring for AI agents. Anchor any SHA-256 hash to Algorand mainnet with ML-DSA-65 (NIST FIPS-204) signatures and receive a self-contained proof bundle verifiable offline — decades from now, without trusting any vendor.
README
PQ Verifiable Archive
Post-quantum tamper-evidence for any signed document. Anchor a SHA-256 hash to Algorand mainnet, get back a self-contained proof bundle signed with ML-DSA-65 (NIST FIPS-204) that anyone can verify offline — decades from now, without trusting any vendor.
Works with any signing tool: DocuSign, HelloSign, Adobe Sign, or a plain PDF. You bring the signed document; this adds the quantum-resistant notarization layer.
Live on Algorand mainnet. Real transactions, real ML-DSA-65 signatures, working offline verifier.
Why this exists
Today's e-signature platforms use RSA or ECDSA. Both are broken by a large-enough quantum computer. NIST finalized ML-DSA (FIPS-204) in 2024 and CNSA 2.0 mandates PQC migration in federal procurement by ~2030–2035. Documents signed today need to be verifiable in 2040.
What's different
| Property | DocuSign / Adobe Sign | PQ Verifiable Archive |
|---|---|---|
| Quantum-resistant | No (RSA / ECDSA) | Yes — ML-DSA-65 + Algorand Falcon-512 state proofs |
| Offline-verifiable | No — requires vendor servers | Yes — 3 of 5 checks need no network at all |
| Vendor-independent | No — trust the platform | Yes — verify against a public permissionless chain |
| PII on-chain | n/a | None — only irreversible SHA-256 hashes are anchored |
| Works with any signer | No | Yes — bring any signed PDF |
| Standards | RSA / ECDSA | NIST FIPS-204, SHA-256, JCS (RFC 8785) |
Use as an MCP tool (AI agents)
Add to claude_desktop_config.json or .claude/settings.json:
{
"mcpServers": {
"pq-verifiable-archive": {
"command": "npx",
"args": ["tsx", "/path/to/pq-verifiable-archive/src/mcp-server.ts"],
"env": {
"ALGORAND_MNEMONIC": "your 25-word mnemonic",
"DOCUSIGN_MLDSA_PUBLIC_KEY": "hex-encoded ML-DSA-65 public key",
"DOCUSIGN_MLDSA_PRIVATE_KEY": "hex-encoded ML-DSA-65 private key",
"DOCUSIGN_KEY_REGISTRATION_TXN_ID": "algorand txn id from npm run register-key"
}
}
}
}
Claude (or any MCP-compatible agent) can then call:
anchor_document— anchor a SHA-256 hash to Algorand, receive a proof bundleverify_bundle— run all 5 post-quantum verification checks on any bundle
Self-hosted MCP is free. You pay only Algorand's network fee (~$0.0002 per anchor).
Hosted API (pay-per-use)
If you don't want to run your own node, call the hosted REST endpoint:
# Without payment — returns 402 with payment instructions
curl -X POST https://pq-verifiable-archive-production.up.railway.app/api/anchor \
-H "Content-Type: application/json" \
-d '{"hash":"<sha256 hex>","envelope_id":"contract-001"}'
# With x402 payment ($0.01 USDC on Algorand)
curl -X POST https://pq-verifiable-archive-production.up.railway.app/api/anchor \
-H "Content-Type: application/json" \
-H "payment-signature: <x402 payment header>" \
-d '{"hash":"<sha256 hex>","envelope_id":"contract-001"}'
Price: $0.01 per anchor. Verification is always free.
OpenAPI spec: /openapi.json — machine-readable for LLMs, tools, and code generators.
Self-hosted quick start
git clone https://github.com/m-reynaldo35/pq-verifiable-archive.git
cd pq-verifiable-archive
npm install
cp .env.example .env # fill in ALGORAND_MNEMONIC + ML-DSA keys
npm run generate-wallet # or use existing wallet
npm run register-key # registers your ML-DSA-65 key on Algorand
npm start # HTTP server on :3000
npm run mcp # MCP server (stdio) for AI agents
Generate a sample document and verify it:
npx tsx scripts/generate-sample-pdf.ts
npm run verify -- --bundle bundles/sample-contract-bundle.json --pdf assets/sample-contract.pdf
Required env vars (see .env.example): ALGORAND_MNEMONIC, DOCUSIGN_MLDSA_PUBLIC_KEY,
DOCUSIGN_MLDSA_PRIVATE_KEY, DOCUSIGN_KEY_REGISTRATION_TXN_ID.
Optional: X402_TREASURY_ADDRESS to enable pay-per-anchor on /api/anchor.
How verification works
The verifier (npm run verify) exits 0 = VALID, 1 = INVALID, 2 = operational error.
| Step | Check | Offline? |
|---|---|---|
| 1 | ML-DSA-65 signature over bundle (NIST FIPS-204) | Yes |
| 2 | SHA-256(PDF) matches bundle.documentHash |
Yes |
| 3 | Merkle proof walks from hash to root | Yes |
| 4 | Algorand txn note contains merkle root | AlgoNode (not DocuSign) |
| 5 | Falcon-512 state proof covers the anchor round | AlgoNode (not DocuSign) |
Who uses this
- HR platforms — offer letters, NDAs, termination agreements with quantum-proof audit trail
- Legal tech — tamper-evident contract archive that survives vendor shutdown
- Healthcare — patient consent forms (no PII on-chain — HIPAA-compatible)
- AI agents — autonomous agents executing agreements need immutable, verifiable receipts
- Anyone signing documents today that need to be verifiable in 2040
Proof bundle (example)
{
"protocol": "pqva/1",
"envelopeId": "contract-2026-001",
"documentHash": "0569e7cb...",
"merkleRoot": "97d5d40b...",
"merkleProof": ["9b1b4a3c...", "34b53407..."],
"algorandTxnId": "QIS2LWKE...",
"algorandRound": 62052659,
"blockTimestamp": "2026-06-11T13:50:11.808Z",
"stateProofRound": 62052864,
"docusignSigners": [
{ "name": "Jordan Avery", "email": "jordan@acme.example", "signedAt": "2026-06-11T13:48:02Z" }
],
"algorithm": "ml-dsa-65",
"mldsaPublicKey": "...",
"signature": "75483d62..."
}
Full schema: docs/architecture.md
Tech stack
| Component | Choice |
|---|---|
| Post-quantum signatures | @noble/post-quantum — ML-DSA-65 (NIST FIPS-204) |
| Blockchain | Algorand mainnet — Falcon-512 state proofs |
| AI agent interface | @modelcontextprotocol/sdk — MCP stdio server |
| Payments | @x402-avm/express — x402 on Algorand (GoPlausible) |
| Merkle trees | merkletreejs — SHA-256 |
| Canonical JSON | canonicalize — JCS, RFC 8785 |
| Runtime | TypeScript + Node.js via tsx |
License
MIT — free to use, self-host, and fork.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。