Veyra

Demo

About

Autonomous AI agents need access to APIs and secrets to be useful, but handing an agent raw API keys is a massive security risk. Furthermore, unchecked agents can spam requests and drain API limits. Veyra solves this by acting as a hardware-backed capability broker based on three pillars of trust. First, users vault their credentials into a physical Ledger Key Ring. The raw keys are never exposed to the internet. Second, when an AI agent needs to execute a task, it routes its request through a Bazantic x402 gateway, paying a crypto micro-toll to prevent spam and monetize the infrastructure. Finally, execution is gated by a World ID Face Auth check on-chain. A verified, live human must approve the specific action before Veyra decrypts the key and executes the task. Agents do the heavy lifting; you hold the keys.

How it's made

Veyra is built with a Next.js frontend, a Node.js backend, and smart contracts deployed on Base Sepolia. Our backend acts as a Universal Hardware Custodian using the Ledger Key Ring CLI (wallet-cli ring). We derive deterministic, collision-resistant hardware slots for every user (keccak256(userAddress) % 2^31), ensuring plaintext secrets never touch the cloud. To expose this local hardware to cloud agents securely, we built an OpenAPI schema and routed a Bazantic x402/MPP Gateway through an ngrok tunnel. This allowed us to demonstrate something entirely new: cloud AI agents paying micro-tolls to encrypt data directly into a local physical hardware enclave. To protect the system from Sybil attacks, we deployed the VeyraRegistry contract on Base Sepolia. The frontend uses World ID's IDKit to generate a ZK proof for human liveness. Our smart contract verifies the proof via the Base Sepolia World ID Router and emits an event. A backend listener catches this event, providing the final human-in-the-loop authorization required to trigger the Ledger decryption command.