n8n's Nod Launches Open-Source Personal Agent

Nod's compelling bet is that trustworthy computer-use agents need both autonomy and visible boundaries. Its research-preview status and local-runtime tradeoffs keep it experimental, but the open-source, model-agnostic approach gives developers unusually direct control.

• –Private virtual-machine execution lets Nod work on websites without requiring an API integration.
• –Users can bring OpenAI, Anthropic, Gemini, OpenRouter, ChatGPT plans, or local models, avoiding vendor lock-in.
• –Approval checks, encrypted credentials, action logs, spending limits, and prompt-injection defenses address the biggest risks in personal agents.
• –The 98% success rate across 678 simulated runs is promising but self-reported and may not reflect messy production websites.
• –Nod must stay on an awake machine, and app connections still route through hosted services by default, limiting its privacy advantage.