ProtocolGovGuardFinGuardExplorerPlaygroundAdoptCloudTrustDocsRequest Pilot
Formally verified · Apache 2.0
26 TLA+ theorems · 35 Alloy facts · view spec
Pre-Execution Trust Layer

Fraud stops before money moves.

Every high-risk action — payments, overrides, approvals — is cryptographically verified before execution. No trust. No assumptions. Only proof.

PROTOCOL CEREMONY · 4 PHASES01EYEobservesBOUND02HANDSHAKEverifiesBOUND03SIGNOFFownsBOUND04COMMITsealsBOUNDEach phase emits a cryptographic event. The chain commits — or refuses.
Automated Tests
3,483
132 test files
TLA+ Theorems
26
TLC 2.19, zero errors
Alloy Facts
35
15 assertions verified
Red Team Cases
85
Cataloged in repo
License
Apache 2.0
Open specification
How EMILIA Works

A control layer between approval and execution.

01 · Intercept

EP sits between approval and execution. Payments, overrides, vendor changes, autonomous AI actions — every high-risk write is gated before it reaches the system of record.

02 · Require Proof

Verified actor identity. Verified authority chain. Policy-pinned action context. One-time nonce. Where policy requires it: a named, accountable human signoff bound to the exact action hash.

03 · Generate Trust Receipt

A signed, Merkle-anchored receipt is produced. Auditor-grade evidence packet at /api/v1/trust-receipts/{id}/evidence. Publicly verifiable with `npm install @emilia-protocol/verify`.

See a real receipt →
Structural Vulnerabilities

Built for approved-looking workflows where ordinary auth fails

Fraud is moving inside valid sessions. Authenticated users, legitimate tools, approved channels — the attack surface is the action itself.

01

Benefit payment redirection

An authorized operator changes a payment destination inside a valid session. No control catches the action itself.

02

Beneficiary and remittance changes

A wire transfer beneficiary is swapped through approved channels. The system sees a legitimate update, not fraud.

03

Privileged infrastructure actions

A production credential is rotated or a deployment is pushed without action-bound authorization. Access was valid.

04

AI agent destructive execution

An agent with broad tool access executes a high-risk action. No human assumed responsibility for the specific operation.

Control Surfaces

Built for the workflows where weak authorization causes real damage

All use cases →

Government Fraud Prevention

Bind identity, authority, and action context before benefit disbursement, procurement approval, or credential issuance.

NIST AI RMFEU AI ACT

Financial Infrastructure Controls

Enforce ceremony-grade authorization on wire transfers, limit changes, account modifications, and privileged treasury actions.

SOX-READYBEC PREVENTION

Enterprise Privileged Actions

Require bound authorization for infrastructure changes, data exports, permission escalations, and production deployments.

ZERO TRUSTPAM LAYER

AI/Agent Execution Governance

Gate autonomous agent actions behind protocol-enforced trust ceremonies before any irreversible real-world execution.

AGENTIC AIHUMAN-IN-LOOP
Implementation Surface

Start anywhere. Go as far as you need.

Zero-dependency verification. Interactive playground. Embeddable trust badges. Everything you need to integrate EP in minutes.

Verify Package

Zero-dependency offline receipt verification. Ed25519 + Merkle proofs. Just math, no EP server required.

npm install @emilia-protocol/verify

Trust Playground

Walk through the EP lifecycle interactively. Create entities, issue receipts, run handshakes — all from one page.

/playground

Trust Explorer

Verify any receipt, proof, or entity. Like Etherscan for trust. Public, transparent, cryptographically verified.

/explorer

Embed Widget

Drop a trust badge on any page. One script tag, one web component. Live data from the EP operator.

<ep-trust-badge />
Initiate Architecture Review

Enforce trust before
high-risk action

See Live ExampleRead the ProtocolRequest Pilot
EMILIA Protocol — Trust Before High-Risk AI Action