Move fast. Stay standing.

The trust infrastructure for the agentic internet.

When your teams ship agents into regulated workflows, the liability is yours. Alignment Cards declare intent, AIP verifies reasoning in flight, CLPI governs the lifecycle on-chain, and Mnemom AEGIS signs the cross-tenant defenses that protect every gateway in the network.

to install
< 5 min
licensed
Apache 2.0
compliance bundles
Audit-ready
Security

A Safe House around every agent.

The Protection Card governs the doors — what gets in, what gets out. The Alignment Card governs the turn — what the agent is permitted to do, and why. Both are signed, versioned, auditable.

Safe House — gateway and observer lanesGATEWAY · REAL-TIMESAFE HOUSE · governed by Protection Cardmid-turngateway logOBSERVER · ASYNCUser/ callerFront doorinbound screeninginjection · hijack · canaryAgent coreAIP loop · nudge / report / blockbound to Alignment CardBack dooroutbound screeningPII · canary · prompt leakUserreceivesAAPpost-hoc verificationdrift · coherence · traceTraceAIP ⊕ AAP · 4-layer cryptoEd25519 · chain · Merkle · ZK(10%)Trust Ratingcomposite 0–1000five components
Protection Card governs the doors. Alignment Card governs the turn — AIP checks every thinking block, mid-turn. AAP verifies post-hoc. Every verdict Ed25519-signed, hash-chained, Merkle-included; ZK-STARK proofs sampled at 10% by default.
Mnemom AEGIS · Protection Network

The first cross-tenant defensive network purpose-built for AI agents.

Safe House is the per-customer perimeter. AEGIS is the network. When one customer's substrate fingerprint shows behavioral deviation, every customer running on that substrate is auto-elevated and a signed Managed Rule lands on every gateway within the propagation SLO.

  • Four checkpoints — front door, back door, inside.autonomy, inside.integrity — each independently set to off, observe, nudge, or enforce.

  • Ed25519-signed Managed Rules with a sub-30s P95 propagation target (first measurements publish 30 days post-GA). Tier-1 and tier-2 rules require dual-control human review under an append-only audit chain.

  • Three signal sources — a 15-persona adversarial arena, customer reports of missed attacks and false alarms, and a cross-tenant aggregator the network can see but no individual customer can.

Supply-chain detection

Per-tenant detection has structurally failed. Substrate fingerprinting catches what Sigstore can't.

Every evaluation is stamped with a substrate fingerprint — provider, model, SDK version, optional lockfile hash. AEGIS attributes anomalies across customers running on the same substrate and propagates a signed Managed Rule before the next agent on the same SDK is hit.

Threats like the Mini Shai-Hulud worm of May 2026 — which compromised 170+ npm packages including Mistral AI's SDK suite and Guardrails AI on PyPI, with valid SLSA-3 attestations on malicious versions — are exactly the cross-tenant pattern AEGIS is designed to detect. We do not replace package-level provenance verification; we are the runtime layer that catches what the supply chain misses.

Intelligence

Know what your agents are actually doing. Before the board asks.

Every agent in your fleet — identified, scored, tracked. Drift surfaced before it becomes an incident. Trust Ratings that travel across OpenAI, Anthropic, Gemini, and local models.

Example
agent/wintermuteagt_01htp8k2m3
A
947/ 1000+2
Updated last 24h
  • ALIGNMENTBound
  • DRIFTStable
  • PROVENANCEC2PA v2.3
  • POLICY BREAKS0 / 1,428
Calm at GA

The IoC feed is empty by design.

The thermometer is calm. The advisory list shows one synthetic post-mortem, clearly labeled. We don't fake activity. The system tells the truth.

Public STIX 2.1 feed/v1/trust/iocs
Trust

Every decision, cryptographically signed. For your board, your auditors, your regulators.

Ed25519 signatures on every verdict. Hash-chained traces, exportable as audit bundles. EU AI Act Article 50 SDK presets ship Article 50-ready disclosures out of the box.

EU AI Act · Article 50
51
days until enforcement

Transparency obligations take effect August 2, 2026. Our Article 50 SDK preset ships Article 50-ready disclosures, logging, and machine-readable content marking.

Compliance posture
EU AI ActEd25519GDPR· readinessHIPAA· readinessSOC 2 Type II· readiness
Compliance

Regulator-ready by construction.

Every governance event is cryptographically signed and append-only audit-chainable. The audit chain is the answer, not a quarterly PDF.

EU AI Act

Articles 10, 12, and Annex IV mapped to the signed governance event chain. Enforcement begins August 2, 2026.

SOC 2 Type II

Readiness program in flight; controls aligned with the AEGIS Managed Rules pipeline and audit chain.

ISO 42001

AI management-system controls mapped to the alignment-card lifecycle and CLPI five-phase governance.

Next step

See the evidence.

A sample coherence report shows how we score a fleet. The Arena shows our detectors holding under live red-team attack.

Next

Bring your own auditor.

Open the customer dashboard. Fetch the IoC feed with curl. Or talk to sales about the enterprise-grade pipeline.

Featured on There's An AI For That