# Mnemom — The trust infrastructure for the agentic internet Move fast. Stay standing. # The trust infrastructure for the agentic internet. Cards declare what agents are. AIP signs how they reason. CLPI signs what they're allowed to do. Mnemom AEGIS defends every transaction across the network — so you can move fast and prove it. [See a sample report](/report/sample)[View Plans](/pricing)[Get Started](/signup) [ sentinel-07 Compliance Agent 847 AA Ed25519SHA-256MerkleAlignment Tap to watch a live simulation → ](/showcase) to install < 5 min licensed Apache 2.0 compliance bundles Audit-ready Security Observatory · Live 04:26:25 UTC Block rate0.42%7d ▼ 0.08 pp Quarantine3 queued24h avg: 4 ZK proof rate10%sampled · default - 04:26:23agent/wintermuteclear·ALIGNMENT HELDL1 - 04:26:21agent/hermes-04quarantine·PROMPT\_INJECTION · koL2 - 04:26:19agent/kurtwood-davisclear·DRIFT STABLEL1 - 04:26:17agent/wintermutesigned·ED25519 · CHAIN OK - 04:26:15agent/parsec-queenreview·UNDECLARED VALUEL2 - 04:26:13agent/hermes-04block·CANARY HIT · P0L1 - 04:26:11agent/navigatorclear·BOUNDED ACTIONL1 Campaigns today: 0Agents protected: 8,402Ed25519 · chain · Merkle · ZK(10%) Gateway running for 0 seconds. The stack ## Five layers. One coherent runtime. Designed from first principles, not assembled from acquisitions. Every layer plays a distinct role; cards are the load-bearing primitive at every step. [ ### AAP Declares intent The Agent Alignment Protocol publishes signed Alignment Cards: declared identity, autonomy bounds, audit commitments. Transparency, not trust. ](/what-we-prove) [ ### AIP Verifies reasoning in flight The Agent Integrity Protocol checkpoints every thinking block before the action executes. Verdict: clear, review needed, or boundary violation. ](/how-it-works) [ ### CLPI Governs and anchors Card Lifecycle & Policy Intelligence — five-phase governance over the cards, with on-chain anchoring on Base L2 and OpenTelemetry observability. ](/governance) [ ### Mnemom AEGIS Signs the cross-tenant defenses Adaptive Enforcement, Governance & Intelligence Substrate. The protection layer of Safe House — screens every transaction at four checkpoints; signs the Managed Rules that propagate across every gateway in the network. ](/learning-network) [ ### Trust Ratings + Coherence Public, portable reputation Cryptographic reputation built from observed behavior across the stack. 0–1000 bond-rated; multi-agent coherence for fleet trust. ](/methodology) 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. [See the Safe House](/security) 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. [See the Protection Network](/protection-network) - Four checkpoints — front door, back door, inside.autonomy, inside.integrity — each independently set to off, observe, nudge, or enforce. - Ed25519-signed Managed Rules with sub-30s P95 propagation. 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 FN/FP reports, 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. [See supply-chain detection](/supply-chain) 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. [See a sample report](/report/sample) 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 [See the advisory list](/trust/advisories)[Fetch the feed](/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 presets ship compliant disclosures out of the box. [What we prove](/what-we-prove) EU AI Act · Article 50 64 days until enforcement Transparency obligations take effect August 2, 2026. Our Article 50 preset ships compliant 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. [See the trust posture](/trust) 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. [See a sample report](/report/sample)[See the Arena](/arena) Next ## Bring your own auditor. Open the customer dashboard. Fetch the IoC feed with curl. Or talk to sales about the enterprise-grade pipeline. [Customer dashboard](/dashboard)[/v1/trust/iocs](https://api.mnemom.ai/v1/trust/iocs)[Contact sales](/contact?reason=enterprise) --- _Source: /index.html · Generated by build-markdown-mirrors.mjs · For agent-readability commitment #4 see https://www.mnemom.ai/for-agents_