Nyxora AI

Nyxora is a Local-First, Zero-Trust AI Middleware designed to help users safely navigate the complexity of DeFi on Arbitrum and other EVM chains. Through natural language, users can execute cross-chain swaps, monitor portfolios, automate recurring DeFi workflows, and interact with on-chain protocols without dealing with fragmented interfaces or complex transaction flows.

Unlike most Web3 AI agents, Nyxora is built around a security-first architecture where the AI reasoning engine never has direct access to private keys. By combining an Isolated Signer Vault, OS-native credential storage, policy-based transaction enforcement, and persistent user memory, Nyxora delivers autonomous on-chain execution without sacrificing self-custody.

The Problem

DeFi remains powerful but difficult to use safely.

Current AI-powered crypto agents introduce several critical problems:

• Insecure Key Management — Many agents require users to store private keys in environment variables, local files, or centralized cloud vaults, creating unnecessary security risks.

• Prompt Injection Risks — AI agents often rely heavily on system prompts and model behavior for security, making them vulnerable to malicious instructions or unexpected model outputs.

• Repetitive User Workflows — Most agents lack meaningful long-term memory, forcing users to repeatedly provide the same instructions and preferences.

• MEV Exposure — Transactions submitted through public mempools can be exposed to front-running and sandwich attacks, leading to poor execution and unnecessary losses.

How Nyxora Solves It

Nyxora addresses these challenges through four core pillars:

1. Zero-Trust Signer Architecture

The AI can plan and reason about transactions, but it cannot directly access private keys.

All cryptographic signing is isolated inside a dedicated Signer Vault process that integrates with the operating system's native credential manager:

• Linux: Secret Service / GNOME Keyring

• macOS: Keychain Access

• Windows: Credential Manager

This separation significantly reduces the impact of compromised prompts, plugins, or AI behavior.

2. Policy-Based Transaction Enforcement

Every transaction passes through a dedicated Policy Engine before reaching the signer.

The Policy Engine enforces hard security constraints such as:

• Maximum transaction size

• Slippage limits

• Asset restrictions

• Approval requirements

This creates a security boundary that operates independently of the LLM.

3. On-Chain AI Kill-Switch (Decentralized Registry)

To provide absolute trustless control, Nyxora embeds a core security layer natively on the Arbitrum Network via the NyxoraAgentRegistry Smart Contract.

Before the local AI can execute any transaction, the system forces an asynchronous RPC check against the blockchain. If a user suspects their local machine is compromised, they can interact directly with the smart contract from any secure device to instantly revoke their agent's status. The local Gateway immediately detects this and terminates the execution thread, making it physically impossible for the AI to transmit the payload.

4. Persistent 4-Layer Memory System

Nyxora continuously learns user preferences and recurring DeFi habits.

Over time, the agent can streamline repetitive workflows such as:

• Scheduled swaps

• Portfolio monitoring

• Position management

• Routine treasury actions

This reduces friction while preserving user control and approval flows.

5. MEV-Aware Transaction Routing

Transactions are routed through MEV-protected infrastructure whenever possible, reducing exposure to front-running and sandwich attacks while improving execution quality.

Vision

To make interacting with Web3 as simple as having a conversation while preserving the principles of self-custody, transparency, and user ownership.

Mission

To build secure, autonomous AI infrastructure that enables everyday users to interact with complex blockchain ecosystems safely, efficiently, and confidently.

Why Nyxora Matters

Most AI agents focus on what the AI can do.

Nyxora focuses on what the AI should never be allowed to do.

By combining autonomous execution with a Zero-Trust architecture, Nyxora enables users to benefit from AI-powered DeFi automation without surrendering control of their assets.

Docs: https://nyxoraai.github.io/Nyxora/

Note for Judges: The smart contract address shown in the pitch video originates from an earlier staging deployment. The final backend infrastructure has since been migrated to Foundry, and the official contract deployed on Arbitrum Sepolia is:

0x6C811587ce9A50f1CF1fE5a42bF2b18a9787318F