n8n 2.x trends: Nod agent and infrastructure shifts

The version number in the title is a misnomer, as the platform has moved well past that point into the 2.x era. The most significant recent development is Nod, a research preview from n8n Labs that functions as a personal agent with its own computer and browser. It runs on macOS 14 or later and Windows 11, handling recurring tasks like chasing overdue invoices or answering client questions directly within your apps. Unlike standard chatbots that only advise, Nod executes these jobs for you, but it asks for permission before sending, paying, or deleting anything you did not explicitly request. This open-source tool uses an Apache 2.0 license and connects to AI providers like OpenAI or Anthropic via your own API keys.

A new agent for personal workflows

Nod is distinct from the core n8n product, which remains focused on shared business processes. It operates as a standalone application that does not require an existing n8n installation, though it can connect to one to build and test workflows. The team at n8n Labs describes it as having "rough edges" because it is still in a research phase, meaning users should expect frequent updates and occasional bugs. Trust is built into the design: emails or web pages cannot grant Nod permissions; only the user can. If you are considering this for your daily operations, note that there is no Linux desktop app yet, though self-hosting on a Linux server is possible. The cost model is straightforward: you pay nothing for the software itself, only for the AI usage through your chosen provider.

Infrastructure shifts and upgrade risks

While Nod captures attention, the underlying infrastructure of n8n is undergoing significant changes that affect deployment strategies. Version 2.43 introduced Browser Use in the Assistant for all users, allowing the AI to read web pages during workflow building, controlled by the setting N8N_INSTANCE_AI_BROWSER_USE_ENABLED. This feature is on by default, so administrators must check it before teams start using live data. More critically, the migration report now labels breaking changes by impact rather than severity, categorizing them as "Upgrade blocked," "Executions fail," "Behavior changes," or "Capability removed." This clarity is vital for the upcoming 3.0 release, which removes nodes like Function and Cron and moves to Docker-only self-hosting.

Self-hosting becomes financially advantageous only when execution volumes exceed roughly 250,000 per month, where cloud bills compress margins. Below that threshold, the labor cost of maintaining servers outweighs the savings. For those already self-hosted, beware of auto-scaling events on AWS; community nodes often stop working after an instance replacement, requiring manual reinstallation. The stable release channel currently points to 2.42.7, while 2.43 builds remain pre-releases. I would wait for the stable tag to move before touching production instances, especially if you rely on legacy Cron schedules that are now running on the durable scheduler. Does your current setup handle the transition to Queue Mode without losing state?