Claude Duet and n8n's Nod are both AI agents that do multi-step chores for you in the background, such as sorting an inbox or drafting replies. The main difference is where they run. Duet runs in the cloud on Anthropic's servers and is a US-only research preview. Nod is a free, open-source agent from n8n that runs in a private virtual machine on your own computer and lets you choose the AI model behind it. For a solo owner, the safe way to start with either one is the same: one small recurring task, a human approval before anything sends or spends, and a hard spending limit set before the first run.
What proactive agents change
A chatbot waits for you to type. An agent like Duet or Nod works from a task you hand it and keeps going without a prompt at every step. Picture a kitchen: the agent is the line cook doing repetitive prep (sorting email, pulling leads), and you are the head chef who plans the menu and checks the plate before it goes out.
Anthropic opened the Claude Duet research preview on October 21, 2026. n8n's Nod arrived a few weeks later; one report dates its launch to November 10, 2026, for Mac and Windows as a research preview. Several labs now sell some version of the same idea (OpenAI calls its agent Dots, Anthropic calls its Duet, Perplexity calls its Teammates, per one X post), so expect this category to keep moving.
Claude Duet (cloud) vs Nod (local)
Claude Duet runs on remote servers, so the heavy computing happens in Anthropic's cloud. It is a waitlisted, US-only research preview for Pro and Max subscribers, which means "open Duet today" is not true for everyone. It uses Anthropic's own models.
Nod runs inside a private virtual machine with its own browser on your computer. Because it is open source, you can plug in the model you prefer, including local models, other providers, or an API key you already pay for. The trade-off is that a local agent only works while your computer is on, and Nod is an early preview with rough edges.
Two name warnings. "Open Duet" is a separate open-source clone of Claude Duet, not an n8n product. And the GitHub project called "claude-duet" is an unrelated community tool.
Can you trust an agent with your passwords?
This is the question that held agents back. According to n8n, Nod keeps your saved logins encrypted on your machine and enters them into websites through private forms the AI model cannot read, so the model sees the resulting page but never the password. We could not find an independent source confirming that design, so treat it as n8n's description and check their documentation before you rely on it. Coverage of the launch also says Nod asks you to type the card security code for each purchase, which adds a second gate on spending.
A local-business example
Say you want to pitch software to a sprinkler repair company. Instead of reading their reviews for three hours (an illustration, not a measured result), you ask the agent to scan the public reviews for recurring complaints. It might find customers frustrated by missed after-hours emergency calls, then draft a pitch built around that one problem.
The key step comes next. The agent stages the draft, and you click send. Anthropic's own safety page says Duet asks before sending sensitive messages or making payments, though it also says those checks can miss things during the preview. So read what you approve.
Two steps to start today
Pick one task and one schedule. Download Nod, or use Claude Duet if you are in the US on a Pro or Max plan and have access. Connect one email inbox and set a single recurring schedule that drafts replies for your most common kinds of message.
Set a hard monthly spending limit first. Before the first run, open your Nod settings or your AI provider's API settings and enter a strict monthly cap. If the agent gets stuck in a loop, the cap stops a surprise bill.
The skill that matters next
If agents handle routine prep work, the skill that matters is judging which tasks are worth handing to a machine. There is a catch for workplaces: entry-level people used to learn by doing those chores, so teams will need a new way to teach that judgment.
FAQ
What is the difference between Claude Duet and n8n Nod?
Claude Duet is a cloud-based agent from Anthropic, currently a US-only, waitlisted research preview for Pro and Max plans. Nod is a free, open-source agent from n8n that runs in a private virtual machine on your own computer and works with the AI model of your choice.
Is n8n Nod really free?
The software is free and open source, but the AI model behind it is not necessarily free. If you connect a paid provider or API key, you pay that provider, which is why a hard monthly spending limit matters.
Is it safe to give an AI agent my passwords?
According to n8n, Nod enters saved logins through private forms so the model never sees them. That claim is from n8n and could not be independently confirmed, so check the documentation, keep human approval on anything that sends or spends, and start with one low-risk task.
What should a small business try first with an AI agent?
Start with one recurring, low-stakes task such as drafting replies to your most common emails. Require your approval before anything goes out, and set a hard spending cap before the first run.
Companion episode:
Want more practical AI breakdowns for small teams? Visit joebuildsai.com.

