AgentProof

Liability Infrastructure for the Agentic Economy

Not just authentication – cryptographic proof of WHO IS LIABLE when AI agents act.

---

The Problem

AI agents are making decisions and taking actions on behalf of humans:

• Transferring money
• Accessing sensitive data
• Managing cloud infrastructure
• Making purchases

But when something goes wrong, who is liable?

Current solutions (Visa TAP, Mastercard Agent Pay) authenticate agents but don't solve liability.

---

The Solution

AgentProof provides Liability Proofs – cryptographic attestations that prove:

1. ✅ A specific human authorized a specific action
2. ✅ Authorization was made via hardware-bound Passkey (unforgeable)
3. ✅ The authorizer explicitly accepts liability
4. ✅ Clear constraints define what's authorized

---

How It Works

┌──────────────┐ ┌─────────────────┐ ┌──────────────────┐
│ Human │ │ AI Agent │ │ Target Service │
│ (Passkey) │ │ │ │ (Bank, API) │
└──────┬───────┘ └────────┬────────┘ └────────┬─────────┘
│ │ │
│ 1. Authorize │ │
│ (Sign with Passkey) │ │
│─────────────────────▶│ │
│ │ │
│ │ 2. Request + Proof │
│ │──────────────────────▶│
│ │ │
│ │ 3. Verify locally │
│ │ (no API call) │
│ │ │
│ │◀──────────────────────│
│ │ 4. Respond │

---

Key Features

Feature	Description
Passkey-Bound	Only device owner can authorize
Self-Verifying	Target APIs verify locally – no latency
Liability Shift	Cryptographic proof of who accepts responsibility
Universal	Works for payments, data access, cloud ops, anything
Decentralized	Trust Mesh – no single point of failure

---

Quick Start

Installation

npm install @agentproof/sdk

Create a Liability Proof

import{AgentProof}from'@agentproof/sdk';constproof=awaitAgentProof.createProof({intent: {action: 'transfer',target: {service: 'api.bank.com',endpoint: '/v1/transfers'},parameters: {amount: 1000,currency: 'USD'}},constraints: {maxAmount: 5000,expiresIn: '5m'}});// Agent includes proof in requestfetch('https://api.bank.com/v1/transfers',{headers: {'X-AgentProof': proof.toHeader()}});

Verify a Proof (Target Service)

import{AgentProof}from'@agentproof/sdk';constresult=awaitAgentProof.verify(req.headers['x-agentproof']);if(result.valid){// Proceed – liability is on the authorizerconsole.log(`Authorized by: ${result.principal.id}`);}else{// Reject – no valid liability proofreturnres.status(401).json({error: result.error});}

---

Architecture

Four Pillars

1. Proof-as-Header – Self-verifying tokens in HTTP headers
2. Trust Mesh – Decentralized P2P trust network
3. Intent DNS – Global, cacheable trust resolution
4. Embedded SDKs – Zero-integration agent runtimes

---

Quick Deployment

Docker (Recommended)

# Clone repository
git clone https://github.com/your-org/agentproof.git
cd agentproof
# Start with Docker Compose
docker compose -f docker-compose.staging.yml up -d --build
# Verify health
curl http://localhost:3000/health

Development

npm install
npm run start:dev
# Run tests
npm run test
npm run test:e2e

Test Coverage

npm run test -- --coverage
# Current: 95% coverage, 319+ tests

---

Documentation

• Protocol Specification
• SDK Reference
• Integration Guide
• Deployment Guide
• Trust Mesh Protocol

---

Comparison

Aspect	Visa TAP	AgentProof
Focus	Authentication	Liability
Scope	Payments only	Universal
Architecture	Centralized	Decentralized
Integration	Merchant work	Embedded in agents
Proof Type	Session-based	Self-verifying

---

License

MIT

---

AgentProof – Prove it. Own it. Trust it.