n8n Releases Nod, a Free Open-Source Personal AI Agent for Mac and Windows

On November 10, 2026, automation company n8n released Nod, a free, open-source personal AI agent for Mac and Windows. The release is a research preview, built by a newly formed team inside the company called n8n Labs, and it runs tasks inside a private sandbox on the user's own computer rather than in the cloud.

The stakes here are straightforward. Most consumer AI agents today live inside vendor-controlled cloud services. Nod takes the opposite bet: the agent, its database, its files, its memory and its saved logins stay on your machine, and you choose which AI provider does the thinking. That's a real design decision with real trade-offs, and it puts n8n into the personal-agent market without requiring anyone to use its workflow platform first.

What exactly is Nod?

Nod is a standalone agent for jobs handled by one person, in their own apps and on websites. It comes with its own browser and can sign in to sites that have no formal integration — no API keys, no official connector. The code is published under the Apache 2.0 license.

You don't need an n8n account to run it, and there's no initial server or cloud-service setup. There is one requirement: you must connect an AI provider. Nod itself is free, but the model usage is yours to pay for, or you can connect an open-source model running locally.

n8n Labs is a new team created for experiments outside the core product. n8n said development of its main product and the n8n agent will continue as planned.

Nod is also deliberately not the n8n AI Assistant and not an n8n workflow. It's a separate thing entirely, though existing n8n users can optionally connect an instance so the agent can work with their workflows. In plain terms: this is n8n trying to own the individual user's desktop, not just the team's automation pipeline.

What can it actually do?

The task list reads like a personal assistant's job description. n8n's examples include:

• Cancelling a subscription and requesting a refund
• Retrieving invoices from supplier portals
• Watching prices and extracting figures from dashboards that lack export functions
• Web and PDF research
• Creating documents, spreadsheets and code in a Linux workspace
• Reading or drafting in Gmail, Calendar, Slack, Notion and Linear

The agent also handles parallelism and scheduling. You can keep one conversation going — in the app, in Telegram or in WhatsApp — while several jobs run at the same time. Recurring work such as morning briefs and price checks can run on schedules. Nod can even propose jobs after reading your connected apps.

That last point matters for adoption. An agent that suggests its own work reduces the cold-start problem that has slowed every personal-agent product so far.

How good is its memory?

This is where n8n makes its most specific claim. Nod retains months of conversation and records how it completed tasks, reusing those notes and revising them when a site changes. It can learn your writing style from your sent Gmail messages. Users can inspect, edit or delete stored information.

The n8n GitHub repository reports a 93.2% LongMemEval score, up from 79%. That figure comes from 500 questions about earlier chats, answered in a fresh session.

The LongMemEval methodology tests five things: information extraction, multi-session reasoning, knowledge updates, temporal reasoning, and abstention when information is unknown. The score therefore concerns long-term memory — not how reliably Nod completes a booking or a refund. That distinction is worth holding onto when evaluating the number.

What stops it from going rogue?

Before Nod sends, pays, posts or deletes anything not covered by the user's request, a separate model-based check asks for confirmation. A web page or an email cannot grant it authority. n8n frames this approval step as a safeguard, and the company is candid about its limits: the check is itself a model, and models make mistakes. It's a safeguard, not a guarantee of correct judgment.

Saved logins and payment cards are encrypted on the machine and entered through private forms the model cannot read. Nod also logs sign-ins, card use, actions in connected apps and model costs, which builds an activity record alongside the approval controls.

Where does your data go?

The data boundaries are the product's headline feature. Nod's database, files, memory and saved logins remain on the machine running it. The selected AI provider receives only the information needed for each task.

The product site adds a nuance: app connections pass through a connection service. Beyond update checks, information goes to n8n only when usage data is enabled or feedback is submitted.

For a company whose core business is cloud-adjacent automation tooling, shipping an agent that sends n8n almost nothing by default is a meaningful positioning move — it targets users who won't put their logins in someone else's cloud.

What does it cost to run?

Nod is free, but the model bills are yours. Supported connections include OpenAI, Anthropic, OpenRouter, Amazon Bedrock, Google Vertex AI and Azure OpenAI. ChatGPT Plus and Pro plans are also supported.

Cost controls include:

• A receipt for each finished task
• A monthly spending limit
• An adjustable check-in once a task passes $5
• An estimated monthly cost for each schedule

Those controls exist because the preview supports scheduled and parallel work drawing on your own model account. Anyone who has left an autonomous agent running against a metered API will understand why a $5 checkpoint exists.

What are the limits of the research preview?

The desktop download requires macOS 14 or later. Windows 11 support is labeled a preview. Self-hosting on a personal server is also available.

There are structural constraints. A laptop installation stops when the laptop sleeps, so scheduled jobs require the machine to remain on. Each installation currently has one owner. And Nod cannot make phone calls.

n8n lists a hosted version and shared team use as future work — both of which would sit in tension with the local-first design, and both of which signal where the company thinks this goes next.

Why does this matter?

Nod gives n8n an open-source route into personal agents without making its workflow platform a prerequisite. It broadens the company's automation focus beyond shared team processes to jobs handled by a single person, whether business or household.

The preview's trade-off is equally clear. Local control and provider choice come with direct model costs, an always-on machine for unattended schedules, and an approval mechanism that can still make mistakes. Whether users accept those trade-offs in exchange for keeping their logins and memory on their own hardware is the question the research preview is built to answer.