n8n Labs Just Released Nod, an Open Source Personal Agent With Its Own Computer

n8n Labs, a new team inside n8n that works on ideas outside the core product, released Nod on 10 November 2026 as a research preview. It is a personal agent with its own computer and browser. You tell it what you need, and it works in your apps and on websites until the job is done.

As an n8n Ambassador I spend most of my time on shared workflows, the kind a team or a business runs on one instance. Nod points at a different job: the errands that belong to one person. That is how n8n frames it too. In their words, a lot of n8n is "multiplayer", and Nod is "for a single player".

I have not run it through a real week yet, so this post is a summary of what n8n published on the announcement, the product page, the docs and the GitHub repo, plus what I think it means for people who already automate with n8n.

There is also a short demo video:

What Nod is

Nod is a standalone desktop app, separate from n8n and from the agents you can build inside n8n. You install it, connect your accounts and ask for something. There is no server or cloud service to set up first, and you do not need an n8n account.

The basics, as listed on the product page and in the README:

macOS 14 or later (Apple silicon or Intel), Windows 11 in preview, or your own server for one person
Free and open source under Apache 2.0
Bring your own model: OpenAI, Anthropic, OpenRouter, or your ChatGPT Plus or Pro plan
Apps connect with one sign-in: Gmail, Calendar, Drive, Slack, Notion
You talk to it in the app, or on Telegram or WhatsApp

n8n says setup takes about ten minutes.

What it actually does

Because Nod has its own computer (a sandbox that runs locally) and its own browser, it can finish a job that a chat assistant would hand back to you as a to-do list. It works in the apps you connect and on websites that have no integration at all.

The examples n8n gives:

Cancel a subscription, turn down the retention offers and ask for a refund
Watch a price and tell you when it drops
Find a hotel that fits your trip and book it once you say go
Sign in to supplier portals each month, download the invoices and file them for your accountant
Pull last week's numbers from a dashboard that has no export and put them in your weekly report

The product page shows three of these as screens. The first is a call brief: every morning Nod reads your threads, your calendar and the client's latest news, then writes a one-page brief for each call.

The second is the follow-up you forgot. Nod finds a question you never answered, here a client's pricing ask in Slack, drafts the reply from your own pricing sheet in Notion, and waits for a yes before sending.

The third is a watcher. You ask it to keep an eye on a page, it checks on a schedule, and it only speaks up when something changes.

It also works ahead of you. It reads your connected apps and suggests jobs it could take off your plate, and it can run several jobs in parallel in the background while you keep one conversation with it. Jobs that repeat, like a morning brief or a price check, can run on a schedule.

Memory

Nod remembers months of conversation. On LongMemEval, a public benchmark for long-term memory, n8n reports a score of 93.2%. When you connect Gmail, it reads how you write so its drafts sound like you, and it keeps notes on how it got a task done, then fixes those notes when a site changes.

You can see, edit or delete anything it keeps about you, and each note shows the message it came from.

The numbers n8n published

From the product page, all measured with GPT-6.1 Sol:

98% task completion across 226 errands in test apps and websites, each run three times, with no actions taken without permission
93.2% on LongMemEval-S, 500 questions about past chats answered in new sessions
0 of 216 on WhisperBench: 108 published attack emails, each tested during an email check and during background monitoring, and none planted a false memory

n8n says the task categories draw on more than 900,000 anonymised production workflows. The method is written up on the benchmarks page. These are n8n's own numbers, so I would read them as a starting point and test on your own errands.

How it handles trust

An agent that can send email and use your card needs guardrails you can check. The ones n8n lists:

The code is open, and every rule it follows is in the repo
It works inside a private virtual machine on your computer, with its own browser. Data, memory and saved logins stay on the machine running it, and the AI provider gets what each task needs
Before it sends, pays, posts or deletes anything you did not ask for, a separate check confirms you asked for it. A web page or an email cannot give it orders
Saved logins and cards are encrypted on your machine and entered through private forms the model cannot read
Every sign-in, card use and app action is logged, along with what each task cost

n8n is upfront that the permission check is itself a model and can make mistakes, which is why Nod asks when your words do not cover an action.

On data: app connections use n8n and Composio by default, and Nod shares usage data with n8n, which you can turn off in Settings. That detail is in the FAQ on the product page, and it is worth knowing before you connect your inbox.

What it costs

Nod is free. You pay your model provider, or you can connect an open-source model running on your own computer. Some early testers found tasks cost more than they expected, and n8n says the latest release brought that down a lot.

The cost guardrails:

Every task shows its cost, and a finished task ends with a receipt
You can set a monthly spending limit
Nod checks in once a task passes $5, and you can change that amount
Each schedule shows its estimated monthly cost

Where it stops, for now

It stops when the laptop running it sleeps. Some testers keep it on an always-on Mac mini or their own server, and a hosted version is on the roadmap
Each installation has one owner
It cannot make phone calls

The roadmap on the product page lists managed hosting, phone calls, one Nod shared by a small team, and making it easier to change Nod in the open code and send changes back.

Where it fits next to n8n

n8n is clear that Nod does not change the direction of n8n, and work on n8n and the n8n agent continues as planned. Their early testers put it in the same spot on their own: the chat assistant is where they think and write, n8n runs the shared processes, and Nod takes the ongoing jobs that are theirs alone.

That split matches how I think about it. If a process is shared, repeatable and needs to be reliable for a team, it belongs in an n8n workflow. If it is my own errand on a website with no API, like a supplier portal or a refund form, a personal agent with a browser is a better fit than building a workflow for something I do once a month. And if you already run n8n, you can connect your instance so Nod can work with your workflows.