Nod

Nod is the standalone personal agent n8n Labs released as a research preview on 10 November 2026. It installs on your computer, has its own local sandboxed virtual machine and browser, can finish work in connected apps and on arbitrary websites, and checks with you before it sends, pays, posts or deletes. The official write-up is Ophir Prusak’s post Meet Nod. The product site is nod.n8n.io, docs are at nod.n8n.io/docs, and the code is at n8n-io/nod. n8n Labs is a new team inside n8n for experiments outside the core product.

In the official blog, a lot of what people build in n8n is multiplayer — shared processes on a shared instance. Nod is for one person, connects that person’s accounts, and does not need an n8n account or a cloud service set up first. It is not the n8n AI Assistant, and n8n says it does not change the direction of n8n itself. Because it has a local computer and browser, official examples include cancelling a subscription and declining retention offers, watching a price, booking a hotel once you agree, signing into supplier portals each month to file invoices, and pulling numbers from a dashboard that has no export. You can talk to it in the app, or on Telegram or WhatsApp; it can run several jobs in the background, and repeating jobs such as a morning brief or a price check can be scheduled.

Trust and cost are central in the post. Nod is open source under Apache 2.0, with its rules in the code. Data, memory and saved logins stay on the machine that runs it; the model provider gets what each task needs. Before Nod sends, pays, posts or deletes something you did not ask for, a separate check confirms that you asked (the post also says this check is itself a model and can be wrong, which is why Nod still asks when your words do not clearly cover an action). Saved passwords and cards are encrypted on the machine and entered through private forms the model cannot read. Every task shows its cost and ends with a receipt; you can set a monthly cap, and Nod checks in once a task passes $5 (that threshold is configurable). You connect your own model — OpenAI, Anthropic, OpenRouter, or a ChatGPT Plus or Pro plan, or a local open-source model. n8n also reports a 93.2% score on LongMemEval; that is the company’s figure, and we have not independently reproduced it. Limits listed in the blog include: a laptop install stops when the machine sleeps (some testers use an always-on Mac mini or their own server); each installation has one owner; it cannot make phone calls yet.

Key links:

• Product: https://nod.n8n.io/ ; blog: https://blog.n8n.io/meet-nod/ ;
• Docs: https://nod.n8n.io/docs/ ; code: https://github.com/n8n-io/nod ;
• This is a standalone app name from n8n Labs. Do not treat it as the n8n workflow product, or as the agents you can already build inside n8n.

Why it matters

Personal agents often either act in the cloud (data leaves your machine) or stop at a to-do list in chat (you still do the last step). Nod tries to finish the work in a local sandbox, publishes the rules as open source, makes “ask first” the default, and shows a per-task receipt. For people who are already comfortable bringing their own model key and want software to handle chores such as bookings or invoices, that is a new name you can actually download and try.

Some pages on the docs site show “last updated” dates earlier than 10 November (the “what is Nod” page has shown 2026-10-29), which means the product had internal or early documentation before the public research preview. This entry is based on the 10 November launch blog. The 93.2% LongMemEval score is official; do not treat it as an independent evaluation without saying so.

Limits

This scan did not find a Nod announcement on the official @n8n_io account in the recent window; the primary sources are the official blog and the product site. A research preview means the UI and data formats can change. A laptop install only runs while the machine is awake, and the permission check can be wrong. Cost depends on the model you connect; early testers found some tasks more expensive than expected, and n8n says a later release cut that substantially — trust the in-app Usage view.

1. Read the official blog https://blog.n8n.io/meet-nod/ and the product page https://nod.n8n.io/ ; use the docs https://nod.n8n.io/docs/ for details ;
2. Download the Mac or Windows build from the product page, or install from source at https://github.com/n8n-io/nod ; connect a model provider and app accounts as the docs describe ;
3. When citing the 93.2% LongMemEval score or “research preview / Apache 2.0,” say the source is the 10 November 2026 official blog, and re-check that page ;
4. Do not treat Nod as the n8n workflow product or as the agents already built inside n8n; this page only restates the official blog, product and docs — re-check the upstream source before citing it.