Prova AgentTrust Protocol

Agent-to-Agent Demo

Any AI agent with TEE-backed identity can request data attestation from the Prova Trust Engine.

Trust Engine Status
Connecting to Trust Engine...

Agent Identity

Each agent gets a DID rooted in its TEE attestation. Identity = code hash + hardware quote.

dstack + W3C DID

Data Provenance

MPC-TLS proves data came from the real source. Notary co-signs without seeing plaintext.

Props + TLSNotary

Attested Computation

AI model weights are pinned to the attestation. Exact model that ran is cryptographically bound.

Props Model Hash
Agent-to-Agent Demo Flow
Simulate an external AI agent requesting attestation from the Prova Trust Engine
1

Agent Identifies Itself

The requesting agent presents its TEE attestation (enclave hash + TDX quote) to the Prova Trust Engine.

2

Submit Data for Attestation

Agent sends raw data (e.g., clinical notes, financial records, legal documents) to the /v1/process endpoint.

3

TEE Processing

The Trust Engine processes data inside the enclave: AI formatting, deterministic validation, anonymization.

4

Attestation Bundle Returned

Agent receives the attested output with full cryptographic proof chain: TDX quote, model hash, MPC-TLS proof, credential destruction.

5

Downstream Verification

Any relying party can independently verify the attestation bundle at /verify — no trust in the requesting agent required.

Multi-Vertical Use Cases

Healthcare

EHR data → FHIR R4 Claims with HIPAA-compliant anonymization

FHIR R4HIPAA 18

Finance

Bank statements → SEC-compliant risk reports with PII stripping

SEC FilingPII (SSN, acct)

Legal

Court records → Structured legal documents with privilege protection

Legal DocAttorney-client

Government

Agency data → Standardized forms with citizen PII protection

Form (1040, W-2)PII (SSN, DOB)

Insurance

Claims databases → ACORD XML with cross-regulation compliance

ACORD XMLHIPAA + PII

Supply Chain

ERP/logistics → Authenticated bills of lading and customs declarations

BOL / CustomsTrade secrets