Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Everyday tasks

Set up Nod, n8n's personal agent

Written November 19, 2026

Nod is a personal agent from n8n Labs that runs on your own computer, with its own sandboxed computer and browser for web work, and asks before it acts for you. It is free and open source, you choose the model it runs on, and n8n puts setup at about ten minutes. This page is the ordered setup: install, model, apps, money rules, first errand.

1. Install the app

Start with the installer that fits your machine, from the n8n-io/nod repository and its install guide. Nod runs on macOS 14 or later, with separate installers for Apple silicon and Intel machines. Windows 11 and Linux x64 are both offered in preview. The desktop app is the normal route: install it on the machine you want Nod to work from, because its files, memory and saved logins stay there.

If you would rather run it yourself, the from-source route works on macOS or Linux. It needs Node.js 24.21 or later, pnpm and Docker. Clone the repository, turn on Corepack, and start it:

git clone https://github.com/n8n-io/nod.git
cd nod
corepack enable
pnpm start

The install-from-source guide covers the extra Linux step for the Vault key and the access link. You do not need n8n itself, or an n8n account, for any of this: Nod is its own app, made by the team behind n8n, and it is not the n8n AI Assistant.

2. Connect a model

Nod runs on a model you choose, and the connect-a-model guide walks through the connection. The provider list covers OpenAI, Anthropic, Gemini, OpenRouter, Bedrock, Vertex AI and Azure, plus more than twenty other providers and your own local server. If you already pay for ChatGPT Plus or Pro, you can sign in with that plan instead of adding an API key. Your provider bills you directly for the model use, and your provider keys never enter the workspace or the browser where Nod does its work.

Choose a capable model here, not the cheapest one listed. The check that decides whether an action matches your request is itself a model making a judgement, so the model behind it sets how much the asking-before-acting behaviour is worth. A stronger model costs more per errand; step 4 is how you keep that trade in view.

3. Connect your apps

With a model connected, connect the apps Nod should work in, following the connect-your-apps guide. The named apps are Gmail, Calendar, Drive, Slack and Notion, and Nod can also work on any website through its own computer and browser, which covers services with no connection of their own. Connect only what your first errands need: each connection widens what Nod can reach, and you can add more later.

Two facts about connections are worth knowing before you grant them. App connections go through a connection service that n8n hosts by default. And while your files, memory and saved logins stay on the machine running Nod, your model provider receives whatever each task needs to do its work.

4. Set the money rules before the first errand

Do this step before you ask Nod to do anything. On API models, every action and sign-in shows in Usage, with the cost of each task, as the spending guide sets out. Set the monthly limit there: Nod stops at the limit you set, so the first month runs inside a ceiling you picked while calm, not one you discover on a bill. If you connected a ChatGPT plan instead of an API key, this Usage view is not where your costs land, so keep the plan's own limits in mind.

5. Hand it a first errand

Pick a first errand of the kind the project itself demonstrates, so you can watch how Nod works before you trust it with more. Three good shapes:

You ask in the app, on Telegram or on WhatsApp, and the conversation stays in one thread across those surfaces. For a wider read on what Nod is and where its limits sit before you install anything, see Nod: n8n's open-source personal agent.

Limits to plan around

Related playbooks

Sources

Back to: Everyday tasks playbooks hub · Also: Nod: n8n's open-source personal agent