Security

Security Architecture

Because Nyxora executes autonomous financial transactions on your behalf, security is our absolute highest priority.

Zero-Trust Policy Engine

Nyxora utilizes a Hard-coded Firewall that sits between the LLM's brain and the Web3 execution layer. This means the AI cannot unilaterally drain your wallet, even if it "hallucinates" or is maliciously prompted.

Execution Boundaries

All actions pass through a signer gatekeeper. The gateway enforces the rules defined in ~/.nyxora/policy.yaml.

• Max USD per Transaction: The agent will refuse to sign any transaction whose estimated USD value exceeds this limit.
• Strict Whitelist Only: If enabled, the agent can only send funds to pre-approved addresses.
• Manual Approval: By default, Nyxora requires explicit manual approval (via the dashboard UI) before broadcasting any state-changing transaction to the blockchain.

Private Key Management

Your private keys are never sent to external APIs (not even OpenAI, Anthropic, or Gemini). They are injected locally during runtime and remain entirely within the sandboxed Node.js environment on your machine.

Where to configure your wallets: Use the DeFi Configuration tab in the dashboard to input private keys or connect a hardware wallet interface (coming soon).

Local Data Isolation

Unlike typical SaaS platforms, Nyxora operates entirely locally.

1. Config Isolation: ~/.nyxora/config.yaml handles generic settings, while ~/.nyxora/config/credentials.yaml handles highly sensitive API keys.
2. No Telemetry: Nyxora does not phone home. Your chat logs, episodic memory (memory.db), and transaction history stay on your hard drive.

Panic Button (Kill Switch)

If you suspect the agent is misbehaving or you need to instantly sever its capabilities, you can use the Wipe All Episodic Memory (Panic Button) in the Settings page. This will immediately reset the agent's learned behaviors and halt active daemon background tasks.