VEX SDK: Verifiable Execution & Silicon Identity 🛡️⚓

The VEX SDK is the official implementation of the Verifiable Execution Protocol. It provides a "black box recorder" for AI agents, making every tool call, decision, and API request mathematically verifiable and cryptographically anchored to hardware.

---

🏛️ Core Relationship

• Provn SDK (provn-sdk / @provncloud/sdk): The cryptographic foundation. It handles identity keys, signing, and low-level data anchoring.
• VEX SDK (provn-vex-sdk / @provnai/vex-sdk): The high-level agent protocol. It handles "Intents," "Authorities," and "Capsules." You use the VEX SDK to build verifiable agents.

graph TD
    A[AI Agent / AI Workflows] --> B[VEX SDK]
    subgraph "VEX Protocol Layer"
        B["VEX SDK (provn-vex-sdk / @provnai/vex-sdk)"]
    end
    B --> C["Provn SDK (provn-sdk / @provncloud/sdk)"]
    subgraph "Cryptographic Foundation"
        C["Base Provn SDK (Low-level crypto)"]
    end
    C --> D[Hardware / Chora Network]

Loading

---

🛡️ Core Pillars

• Verifiable Execution (VEX): Generates tamper-proof Evidence Capsules (.capsule) for every agent action.
• Silicon Identity (Attest): Natively integrates hardware-rooted trust (TPM 2.0, Secure Enclaves) to bind actions to specific physical machines.
• Protocol Parity: Guaranteed bit-for-bit binary parity across Python and TypeScript implementations.
• Cognitive Routing: Seamlessly intercept and secure agent tool calls via high-level decorators and middleware.

---

🚀 Quick Start

Python

pip install provn-vex-sdk

from provn_vex_sdk import vex_secured

@vex_secured(intent="Critical system update")
async def perform_update(params: dict):
    # Automatically generates and verifies proof before execution
    return await internal_api.update(params)

TypeScript

npm install @provnai/vex-sdk

import { vexMiddleware } from '@provnai/vex-sdk';

// Secure Vercel AI SDK or generic tool loops
const securedTools = vexMiddleware({ 
  identityKey: process.env.VEX_KEY!, 
  vanguardUrl: process.env.VEX_VANGUARD_URL! 
});

---

⚙️ How it works

VEX builds a .capsule artifact for every action. This envelope contains:

1. The Intent: The objective context of the action.
2. The Authority: Permission and governance signals (VEX/CHORA).
3. Silicon Identity (Attest): Proof the hardware is genuine and secure.
4. The Witness: A timestamped signature from the Chora network.

graph LR
    A[AI Agent] --> B[VEX SDK]
    B --> C{Verified?}
    C -->|Yes| D[Go ahead!]
    C -->|Hold| E[Escalate/AEM]
    C -->|No| F[Halt Execution]

Loading

---

🏛️ Monorepo Structure

• /python: Official Python SDK (pip install provn-vex-sdk).
• /typescript: Official TypeScript SDK (npm install @provnai/vex-sdk).
• /docs: Technical specifications, Silicon Identity (Attest) guide, and audit reports.

---

🤝 Contributing & Security

We built VEX to be open and auditable.

• See CONTRIBUTING.md to help us build the future of AI trust.
• See SECURITY.md to report vulnerabilities privately.

---

Developed by ProvnAI as part of the ARIA Scaling Trust programme. 🛡️⚓🚀