18/18 E2E TESTS PASSING — LIVE WITH OPENAI

Every Agent Step.
Governed. Signed. Replayable.

GRAFOMEM is the compliance layer your AI agents are missing. Hash-chained execution receipts, signed decision trails, and GDPR erasure certificates — out of the box.

18/18
E2E Tests Passing
3
Hash-Chained Receipts
INTACT
Chain Verification
1,166
Tokens Governed
12
Policy Evaluations

Three Pillars of Governed AI

Memory, Governance, and Attestation — the complete stack for compliant AI agent deployment.

🧠

Verifiable Memory

GMP-conformant memory stores with PostgreSQL + pgvector. Every write, retrieve, delete, and supersede is audit-logged with full provenance.

  • 384-dim HNSW vector search
  • Ed25519 content signatures
  • Multi-store per tenant
  • Semantic + keyword retrieval
🛡️

Policy-as-Code Governance

PDP/PEP architecture with a separated Policy Engine and Governance Gateway. Rate limits, PII guards, and human-in-the-loop escalation — enforced before every inference.

  • Stateless policy evaluation
  • Append-only evidence logs
  • PII detection + redaction
  • HITL escalation gates
🔗

Cryptographic Attestation

Every agent step produces a hash-chained execution receipt. Every decision is Ed25519-signed. Every deletion generates a GDPR erasure certificate. Tamper-evident by construction.

  • BLAKE2b-256 hash chains
  • Deterministic replay engine
  • Signed erasure certificates
  • One-click regulatory reports
STEP 1 — Researcher ✓ SIGNED
receipt: 7b27ec01... | input_hash: a3f8d2...
STEP 2 — Writer ✓ SIGNED
receipt: 9e41f5ab... | prev_chain: 7b27ec...
STEP 3 — Reviewer ✓ SIGNED
receipt: c4d89a12... | prev_chain: 9e41f5...
🔗 CHAIN INTACT 3/3 steps verified

Conformance &
Attestation

GRAFOMEM doesn't just run agents — it attests every step. Five cryptographic proof types, aligned with real regulations.

📋
Execution Attestation
Hash-chained receipts prove this workflow ran these steps in this order.
→ EU AI Act Art. 12 (Logging)
🔏
Inference Provenance
Ed25519-signed decision records with full input/output context.
→ EU AI Act Art. 14 (Oversight)
🗑️
Deletion Attestation
Signed erasure certificates with content hash. Prove data was deleted.
→ GDPR Art. 17 (Right to Erasure)
🔄
Decision Reproducibility
Replay any decision with the same inputs. Measure confidence drift.
→ DORA Art. 6 (ICT Resilience)
📊
Policy Enforcement
Append-only governance evaluation logs. Every gate recorded.
→ ISO 42001 (AI Management)

7-Layer Governance Stack

Built on the open GMP specification (MIT, v0.2.0). Every layer adds a governance guarantee that compounds into compliance-grade AI.

Layer 1 — Memory
GMP-conformant stores · PostgreSQL + pgvector · HNSW 384-dim · Write / Retrieve / Delete / Supersede
Layer 2 — Decision Trail
Ed25519-signed inference logs · Full context capture · Retrieved facts + prompts + outputs
Layer 3 — Erasure Proof
GDPR Article 17 certificates · Content hash attestation · Signed deletion proof
Layer 4 — Governance
PDP/PEP architecture · Policy Engine (stateless) · Evidence Collector (append-only) · HITL gates
Layer 5 — Reports
One-click EU AI Act / GDPR / DORA compliance packages · Risk assessments · Audit bundles
Layer 6 — Orchestrator
8-step governed execution loop · Sequential / Supervisor / Round-robin · BYOM LLM registry
Layer 7 — Attestation
BLAKE2b hash chains · Execution receipts · Deterministic replay · Tamper detection

Start Free. Scale with Compliance.

Every plan includes full governance. Enterprise adds HITL, SSO, and dedicated infrastructure.

Starter
$0
Free forever · 1,000 operations/mo
  • 1 memory store
  • Decision trail logging
  • Governance gateway
  • Erasure certificates
  • Community support
Get Started Free
Enterprise
Custom
Contact sales · Unlimited
  • Everything in Pro
  • HITL escalation flows
  • SSO / SAML integration
  • Dedicated infrastructure
  • Custom compliance reports
  • SLA + dedicated support
Contact Sales

Ready to Govern
Your AI Agents?

Deploy with confidence. Every decision logged. Every step attested. Every deletion proven.