Let Nod by n8n Handle the Chores You Keep Putting Off

What you'll get

Nod is an open-source personal assistant from n8n, currently in research preview. It runs on your own Mac or Windows computer and has its own small virtual machine, so it can browse the web, read email, check your calendar, fill in forms, and handle chores like chasing unpaid invoices, preparing for meetings, or canceling a free trial you forgot about. This guide walks you through installing it, connecting a model and your apps, handing it a first task, and gives you a few task phrasings you can copy.

You don't need to write code or build an n8n workflow. You only do three things: say what result you want, nod or shake your head when it asks, and review what it produced. Before it sends anything or pays for something, it stops and asks you.

The full path looks like this: install the desktop app, connect a model, link your email and calendar if you need them, describe the task and its limits in one sentence, and approve it when it's about to send or pay.

Step 1: Download and install

Download the desktop app from Nod's GitHub releases page and install it like any other app:

Computer	Requirements	Download size
Mac	macOS 14 or later. Apple silicon and Intel both work	About 1.8 GB
Windows	Windows 11 (preview). Needs WSL 2, plus 20 GB or more of empty disk space	About 2.1 GB

When you open it for the first time, pick Get started. It will set up its own virtual machine and then walk you through a few settings. On Windows, if it asks you to install WSL, go ahead and accept. If it needs a restart, restart. Your settings will still be there when you come back.

If you hit an installer message you don't understand, paste it to your AI agent:

I'm installing Nod, the open-source personal assistant desktop app from n8n, and I got this message: [paste the exact message here]
Please explain in plain language what it means and which button I should click. If I need to run a command in the terminal, first tell me what the command does, and only run it after I agree.

Step 2: Connect a brain

Nod doesn't come with a model, so you need to connect one. You can change it anytime under Settings → Models:

If you have ChatGPT Plus or Pro: choose OpenAI, then choose sign in with ChatGPT. Approve access in the web page that opens. It uses the quota in your subscription.
If you want to try it for free: grab a Gemini API key in Google AI Studio; the Gemini 3.8 Flash model comes with free usage.
If you want zero cost: connect a local model, such as one running through Ollama on your machine.
If you want to use Claude: you need an Anthropic API key. You can't sign in with a Claude subscription.

Step 3: Connect only the apps you need

Nod can connect to Gmail, Google Calendar, Slack, Notion, Linear, and more. It can also keep a conversation going with you in Telegram or WhatsApp. For your first run, connect one or two apps, like email and calendar, and add more once you're comfortable.

Website logins and card details are stored in its vault. The model never sees the raw values.

Step 4: Hand over your first chore

When you give it a task, say three things: the result you want, which accounts and websites it can use, and where it must stop and ask you. Here are three common requests. Swap the placeholders in brackets and you're set:

New accounts get 10 free credits, plus 1 more for each daily check-in. Spend 1 credit to unlock and it stays open for you.

Here's why this phrasing works: Nod decides whether it can act right away based on your exact words. If you say "read and draft only," it won't send anything. If you say "don't send until I say so," it will wait for you. The clearer your limits, the fewer questions it asks and the less it does on its own.

Step 5: Approve when it asks

Every action that could affect the outside world, like sending an email, submitting a form, making a payment, or changing a setting, goes through a checker on the server. The checker compares the action with what you said and decides whether you approved it:

Covered by your words: If you say "book a table for two tonight at 8," it will choose a time, fill in the names, and confirm the reservation without stopping.
Beyond your words: If it wants to pay a deposit, switch to a different restaurant, or message someone else, it stops and asks you.

Everything it does shows up on the Usage page: each send, login, and card charge, with a note about which of your messages allowed it. You can also set a monthly spending cap.

The screenshots above come from Nod's official docs and its GitHub repo.

Want it to watch things on a schedule?

Just say something like "every weekday at 8 a.m., check my schedule for today and warn me about any conflicts" in the chat, and it will create a scheduled task. One catch: if it's running on your own computer, scheduled tasks won't run while the computer is asleep or while Nod is closed. To keep them running around the clock, you'll need to deploy it on a server you control.

FAQ

How do I use Nod by n8n? Install the desktop app, connect a model (a ChatGPT subscription, a free Gemini key, or a local model all work), link your email and calendar, and then tell it in the chat what result you want and where it should stop and ask you.

Is Nod free? Nod itself is free and open source under the Apache 2.0 license. For the model, you can use an existing ChatGPT Plus or Pro subscription, pay per use through an API, or run a local model at no cost.

What's the difference between Nod and n8n? n8n is a tool where you drag nodes around to build automation workflows yourself. Nod is a personal assistant from the n8n team. You give it tasks in a chat, and it plans the steps on its own, so you don't build any workflow.

Is Nod safe? Could it send emails by mistake? Any action that sends something outward, makes a payment, or changes a setting has to pass the server-side check first. Anything beyond your original words gets stopped and sent back to you. That said, the checker is also an AI model and can make mistakes, so look over anything important before you approve it.

Nod won't open or doesn't work on my Mac. What should I do? First, confirm that you're on macOS 14 or later and that you downloaded the build for your chip (Apple silicon or Intel). The first time you open it, macOS will warn you that the app came from the internet. Choose Open. If it still fails, paste the exact error message to your AI agent and ask it to help you troubleshoot.

Watch out

It's still a research preview: Versions change quickly, and the interface and features may shift.
Basic usage data is sent by default: To turn that off, go to Settings → Privacy & data.
The Windows version is a preview: It needs WSL 2 and hardware virtualization, so older computers may not install it.
Start small: For the first few days, don't hand it payments or important outgoing emails. Let it read and draft first, and give it more room once you trust how it works.