Source-only acquisition opportunity

Evidence that outlives the system that created it.

Post-quantum cryptographic evidence infrastructure for AI content, audit and risk workflows. 132,000+ lines of Rust, Python and Web code. 31 crates. 4 SDKs. Built, tested, and available for acquisition. No customers. No revenue. Just the engineering.

132K+ Lines of Code
31 Rust Crates
2,775+ Tests Passing
SLSA L3 Build Attestation
terminal
$ trai sign --text "Q2 financial report generated by AI assistant"
SHA-256 hash computed
Ed25519 + ML-DSA-65 signed
SCITT receipt anchored (RFC 9943)
eIDAS timestamp via Actalis Italia
Cert ID cert_9f8a2b1d5f3e9c7a
Verify trai.apohara.io/verify/cert_9f8a...
$
The code speaks for itself

Not a concept deck. Real engineering.

Every line backed by a test. Every module auditable. Every claim verifiable against the source repository.

Rust — Core Cryptography crates/tl-evidence/src/sign.rs
pub async fn sign_cose(
    payload: &[u8],
    key: &SigningKey<Ed25519>,
) -> Result<CoseSign1> // Ed25519 + ML-DSA-65 hybrid signing
    // FIPS 204 aligned · COSE_Sign1 / RFC 9052
    // 0 todo!() · 0 unimplemented!()
    // #![deny(clippy::unwrap_used)]
    let sig = key.sign(payload);
    let cose = build_cose_envelope(payload, sig);
    Ok(cose)
Python — Control Plane services/control_plane/app/api/notarize.py
@router.post("/v1/notarize")
async def notarize(
    req: NotarizeRequest,
    db: AsyncSession = Depends(get_db),
):
    # Multi-tenant · JWT org resolution
    # Append-only audit tables
    # HSM failover · 5 providers
    evidence = await create_evidence(
        text=req.text,
        framework=req.framework,
    )
    return EvidenceResponse(certificate_id=evidence.id)
CLI — Developer Experience terminal
# Sign content and generate evidence certificate
$ trai sign --text "$(cat report.txt)" \
    --framework eu-ai-act-art50 \
    --disclosure "Generated by AI system v2.1"

 SHA-256 hash computed (0.3ms)
 Ed25519 + ML-DSA-65 signed (0.8ms)
 SCITT receipt anchored
 eIDAS timestamp via Actalis Italia

Certificate: cert_9f8a2b1d5f3e9c7a
Verify: trai.apohara.io/verify/cert_9f8a...
Engineering moat

The part that is expensive to rebuild is already built.

MODULE STATE LOC REVIEW
Rust crates (31 workspace)READY83,697/crates/
FastAPI control planeREADY26,188/services/
Frontend (Astro + Svelte)READY19,272/web/
SDKs (Python, TS, Go, WASM)READY1,887/sdk/
MCP Server (37 tools)READY4,928/crates/tl-mcp-server/
F* formal proofsDOCUMENTED/formal/
CI/CD (19 workflows)SLSA L3/.github/
Docker Compose productionREADY/docker-compose.prod.yml
PostgreSQL + AlembicBASELINE/services/control_plane/
Adversarial testing (6 fw)READY/fuzz/
EU Trust List refreshPARTIAL/docs/KNOWN_LIMITATIONS.md
Live Evidence Pipeline

Do not take the claim. Test the chain.

Every hash, every signature, every timestamp runs in your browser via Web Crypto API. No mock data. No fake latency. Real cryptographic primitives. The SCITT receipt uses a verified sample artifact. Tamper detection demonstrates signature invalidation.

EU AI Act Art. 50 enforcement
--- Fine: €15M or 3% of global annual turnover (Art. 99)
1
SHA-256 Hashing Cryptographic digest of content via Web Crypto API
Computed locally
2
Ed25519 Signing Browser-generated signing keypair via Web Crypto API
Computed locally
3
BLAKE3 Hash Chain Content-addressed chain entry via WASM compute_canonical_hash
Computed locally
4
SCITT Receipt Transparency log receipt (RFC 9943) with Merkle inclusion proof
Verified sample artifact
5
Tamper Verification Cross-check signature against original content hash
Computed locally
Evidence Substrate

What you acquire

Four integrated components. Complete source code. No vendor lock-in. No subscription required.

MCP Tools

36 tools across 9 modules

Model Context Protocol integration for AI coding assistants. Each tool is independently callable and backed by Rust cryptographic primitives.

ClaudeCursorCodex
  • Signing, verification, timestamping, SCITT anchoring
  • Compliance mapper endpoints (Art 50, DORA, PLD)
  • PDF certificate generation and audit trail
  • Offline WASM verification bundle
View in action

4 SDKs + WASM

108KB browser bundle

Self-hosted cryptographic compliance SDKs. No SaaS dependency. Python, TypeScript, Go, and a 108KB WASM bundle for browser-native verification.

PythonTypeScriptGoWASM
  • Sign, verify, timestamp, anchor — any language
  • Offline verification without network calls
  • KMS integration (AWS, GCP, Azure)
  • CLI for CI/CD pipeline integration
View SDK docs

PDF Certificate

3-tier disclosure format

Court-grade evidence certificates with QR verification, embedded SCITT receipts, and qualified eIDAS timestamps from Actalis Italia.

A4 FormatQR CodeeIDAS
  • L1 summary / L2 full CWT / L3 verification steps
  • Embedded COSE_Sign1 envelope with public key
  • RFC 3161 timestamp token from qualified TSP
  • Merkle inclusion proof for SCITT anchoring
Download sample

Agent Console

Real-time compliance dashboard

Monitor AI agent compliance status, evidence chains, and verification outcomes. Integrates with existing SIEM and GRC workflows.

DashboardSIEMGRC
  • Agent registration and key management
  • Evidence chain visualization
  • Bulk verification and audit export
  • Webhook integration for compliance events
Open console
Category Comparison

How the approaches differ

Three categories of compliance tooling. Each serves a different need. TRAI is a source asset for organizations that want to own their evidence substrate.

Need Dashboard / GRC Vendor-managed TRAI Source Asset Buyer-operated
Framework mapping
Vendor-managed compliance databases
Technical mapping (buyer-operated)
Cryptographic evidence per output
~ Varies by vendor; most use API calls, not local crypto
Ed25519 + ML-DSA-65 hybrid signatures computed locally
Ownership model
Subscription / vendor lock-in
Buyer-operated source code; full IP assignment
Post-quantum readiness
Roadmap dependent; most lack ML-DSA-65
Ed25519 + ML-DSA-65 hybrid (FIPS 204) implemented today
Production operation
Vendor-managed infrastructure
Buyer provisions own infrastructure and keys
Offline verification
Requires vendor platform connectivity
4 SDKs + WASM; fully offline capable
SCITT transparency anchoring
Not implemented in most compliance tools
RFC 9943 receipt format with Merkle inclusion proofs
eIDAS-qualified timestamps
Some offer timestamping; rarely qualified TSP
RFC 3161 format; qualified credentials transferred with sale

Comparison reflects category-level capabilities. Dashboard/GRC tools excel at workflow automation and regulatory database management. TRAI focuses on cryptographic evidence generation at the output level. The approaches are complementary, not mutually exclusive.

Due Diligence

Honest state of the codebase

What's production-ready, what needs buyer infrastructure, and known gaps. Transparency for informed due diligence.

Production Ready

  • COSE_Sign1 signing — Ed25519 + ML-DSA-65 hybrid (FIPS 204)
  • 3 Jan-2026 CVEs patched via ml-dsa >= 0.1.0-rc.5
  • RFC 3161 timestamp — format implemented (qualified TSP requires buyer credentials)
  • RFC 9943 SCITT receipt format — format implemented (production registry requires buyer operation)
  • BLAKE3 append-only hash chain (P4.6 encapsulated)
  • Multi-tenant SQL isolation — cross-tenant returns 404, never 403
  • FastAPI control plane with 18+ endpoints, JWT-gated
  • SLSA Build L3 attestation (Sigstore + Rekor)
  • OpenSSF Best Practices self-assessment complete
  • cargo test --workspace — 2,775+ pass, 0 fail
  • F* proof of COSE_Sign1 Sig_structure (RFC 9052)
  • Adversarial testing: 6 frameworks
  • POST /v1/rebut — PLD Art. 10 evidence bundle (COSE_Sign1 + RFC 3161 + SCITT)
  • Ed25519 + ML-DSA-65 throughput bench (criterion, software baseline)
  • EU Trust List auto-refresh (24h LOTL background fetch, cached to disk)
  • Caddyfile TLS termination (ACME, CSP, HSTS, reverse proxy)
  • docker-compose.prod.yml (PostgreSQL 16 + Alembic migrations)

Needs Buyer Infrastructure

  • PostgreSQL persistence — docker-compose.prod.yml included; buyer provisions RDS / Supabase for production scale
  • HSM signing at prod volume — software throughput baselined; HSM backend requires buyer KMS and load testing
  • Co-maintainer #1 expected Q3 2026 — current bus factor 1
  • TLS termination — Caddyfile included; buyer points DNS and runs caddy run
  • Multi-region redundancy — single-region only today
  • SD-JWT selective disclosure — deferred, not implemented
  • C2PA content provenance — stub only, not implemented
  • EU Trust List — 24h auto-refresh wired; LOTL cache integrated with QTSP validation is pending deeper work
  • HSM-backed ML-DSA-65 signing at production concurrency — software bench only, needs HSM integration testing

Roadmap

  • OpenSSF Silver — code artifacts prepared (post-acquisition buyer submission, not included in sale)
  • OpenSSF Gold — target Q4 2026 (post-acquisition buyer option, not included in sale)
  • SOC 2 Type II — target Q1 2027 (post-acquisition buyer option, not included in sale)
  • ISO/IEC 42001:2023 — Stage-1 audit Q2 2028 (post-acquisition buyer option, not included in sale)
  • C2PA v3.0 — wait for ratification (v2.4 current)
  • PQC-only Ed25519 retirement — hybrid SOTA path until 2028
Acquire the proof layer

Acquire the proof layer before you build it.

132,000+ lines of code. 31 Rust crates. 4 SDKs. 2,775+ tests. SLSA L3 build attestation. F* formal proofs. Full IP assignment. No subscription. No vendor lock-in. A buyer should be able to understand the core system in an afternoon, operate a demo in a day, and decide what to productionize next.

Ed25519 + ML-DSA-65 hybrid
RFC 9943 SCITT receipts
RFC 3161 eIDAS timestamps
2,775+ tests passing

TRAI provides technical evidence generation capabilities. It does not constitute legal advice, regulatory certification, or a substitute for independent compliance review. Buyers must validate regulatory applicability in their jurisdiction. Built to support Article 50 transparency workflows — not a substitute for legal compliance assessment.