Measured. Not promised.

Every number on this page is a test result. Every claim has a proof artifact. We don't publish benchmarks we can't reproduce.

Current verified numbers

900+
Tests Passing
Across Node.js and Python stacks
<8ms
Receipt Latency
Production SLA, measured
117/117
Attacks Blocked
Adversarial test suite
82 → 0
Breaches
Pre-hardening → post-hardening
36
API Endpoints
OpenAPI 3.1 documented
29+
USPTO Filings
3 non-provisional + 26 provisional

Test coverage by layer

Every governance layer is tested independently. These numbers come from the actual test harnesses, not marketing estimates.

173
Equivalence Engine

42 temporal + 39 regulatory + 92 compositional drift

111+
Trust Scoring

Scoring accuracy, authorization gates, L1 receipt wiring

59
Veto Registry

HARD/SOFT vetoes, scoping, fail-closed, input sanitization

65+
Weight Profiles

Tenant weights, epoch binding, validation, immutability

38+
Chain Anchoring

Merkle tree building, batch anchoring, proof generation, verification

46
Receipt Chain (L1)

Chain linking, counter integrity, isolation, full chain audit

213
CDT Engine (Python)

Kernel sovereignty, fail-closed, boundary proof, gate proxy

44
IDV System

Auth, chain, export, falsifiability, governance, verification

Red team results

Three independent assessment passes: architectural red team, adversarial attack suite, and white-box code scan.

15
Architectural Findings
14 fixed, 1 documented
117
Attack Vectors Tested
117 blocked (0 breached)
99
White-Box Findings
All CRITICAL and HIGH remediated
0
Layer Isolation Violations
No upward writes detected
VERDICT: HARDENED

What a receipt proves

Every governed operation produces a cryptographic receipt. Each receipt is independently verifiable and contains:

What was decided (ALLOW or DENY)
When it was decided (tamper-evident timestamp)
What policy was in effect (cryptographic hash binding)
That the policy didn't change between authorization and execution
Who had authority to make the decision
That the chain of receipts is unbroken and sequential

What happens on denial

The system fails closed. A denial is not silent — it produces as much evidence as an approval.

Execution is blocked (fail-closed, not fail-open)
A DENY receipt is emitted with the specific reason
The DENY receipt is chain-linked to the prior receipt
The agent's trust score is updated to reflect the denial
The snapshot is consumed (no replay possible)

Publicly demonstrated

✓Cryptographic receipts at the execution boundary
✓Fail-closed denial on invalid or stale authority
✓Drift detection between authorization and execution
✓Continuity behavior preserved across governed operation
✓Evidence suitable for later review and offline verification
✓Test counts per governance layer
✓Red team and adversarial result summaries
✓Benchmark latency ranges
✓Live deployment URLs

Available under NDA

●Receipt schema and chaining mechanics
●Continuity binding and field-level details
●Trust scoring factors, weights, and thresholds
●Deployment topology and integration architecture
●Cryptographic implementation specifics
●Full adversarial test suite with attack vectors
●Solidity contract source and deployment details

See it live.

Request a live verification walkthrough. We'll run the full evaluate → commit → drift → receipt chain audit in under 60 seconds.

Request DemoTechnical Overview