n8n Labs released Nod on 10 November, a free, open-source personal agent with its own computer and browser that works in apps and on websites and uses the AI provider you choose, according to n8n’s blog. The project’s GitHub README adds a test of its permission check: in 240 test decisions, GPT-6.1 Sol and Claude Opus 5.5 “got every one right”, while “Claude Haiku 5.5 let 8 unsafe actions through”.
Eight of 240 is 3.3%, by our arithmetic. On Anthropic’s pricing page, Haiku 5.5 costs $0.10 per million input tokens for prompts up to 100,000 tokens, one fortieth of Opus 5.5’s $4. Nod’s documentation says that “by default, the checker uses your selected model at its lowest reasoning setting”.
Business Pill · Prompt Injection Explained
An evergreen Business Engineer explainer on prompt injection: when text a model reads gives the orders. It teaches the general idea only and does not discuss Nod.
The key insight: As we read it, n8n has made the agent free and left the model to the user, and its own test shows that the model choice also sets how well the permission check holds: Claude Haiku 5.5 let 8 of 240 unsafe actions through, 3.3% by our arithmetic, while Opus 5.5 and GPT-6.1 Sol let none through.
The blog calls Nod a research preview from n8n Labs, “a new team within n8n that experiments with ideas outside our core product”. Nod is a standalone app: you do not need an n8n account, and it runs on Mac and Windows or on your own server. You talk to it in the app, on Telegram or on WhatsApp.
The jobs n8n lists include cancelling a subscription and asking for a refund, watching a price, booking a hotel once you say go, downloading invoices from supplier portals, and pulling numbers from a dashboard that cannot export them. The blog says Nod works “on websites that have no integration at all”.
Nod is licensed under Apache 2.0. n8n’s own main product is published under its Sustainable Use License, which allows use “only for your own internal business purposes or for non-commercial or personal use”. The Nod repository showed 81 stars and 6 forks at about 10:00 UTC on 11 November; n8n’s main repository showed 210,121 stars.
n8n announced Nod on X at 08:19 UTC on 11 November, and Wall St Engine relayed it at 09:22 UTC. Block released another free, self-hostable agent this week; our piece on Block’s Buzz 1.0 covers its pricing.
Unsafe actions let through in 240 test decisions of Nod’s permission check, by model, as published in the n8n-io/nod README: GPT-6.1 Sol 0, Claude Opus 5.5 0, Claude Haiku 5.5 8. API input prices from Anthropic’s pricing page: Opus 5.5 $4 and Haiku 5.5 $0.10 per million tokens. Our arithmetic: 8 of 240 is 3.3%.
What Was Posted on X
Meet Nod, your own open-source personal agent from n8n Labs
Ask it to prep your calls, collect invoices or cancel a subscription
It runs on your computer with its own browser to get the job done across apps and websites, using the AI provider you choose๐
n8n’s own announcement on X at 08:19 UTC on 11 November, a day after its blog post: Nod is an open-source personal agent from n8n Labs that runs on your computer with its own browser, using the AI provider you choose.
n8n just launched Nod, a free, open-source AI agent that works across apps and websites, even without APIs
It runs on Mac or Windows with its own browser to handle everything from emails and calendars to invoices and subscription
Wall St Engine relayed the launch at 09:22 UTC, noting that users pick the AI provider and that usage costs extra. The piece’s figures come from n8n’s blog, README and docs, not from this post.
How the Permission Check Works
The blog describes the rule: “Before Nod sends, pays, posts or deletes anything you didn’t ask for, a separate check confirms you asked for it. A web page or an email can’t give Nod orders.” It adds that “the permission check is itself a model and can make mistakes”.
Nod’s docs say a server setting can choose a small model or decision models for the checker instead of the default. Separately, “background work, such as memory upkeep, uses a smaller model from the same connection when it has one, such as gpt-6-luna or claude-haiku-5-5”.
The README’s advice next to the 240-decision result is “Pick a capable model.” It does not say when the 240 decisions were run or list them. Our piece on agent risk in company filings covers how listed companies describe agents taking unintended actions.
What It Costs
The blog says Nod is free: “You pay your model provider for the AI you use, or you can connect an open-source model running on your own computer.” It also says “some early testers found tasks cost more than they expected”, and that Nod checks in once a task passes $5, with a monthly spending limit you set.
Nod can sign in with a ChatGPT Plus or Pro plan, its README says. OpenAI’s pricing page lists Plus at $20 a month and Pro from $100 a month; Pro lists “Dot, your always-on agent”. Nod’s models page connects Anthropic with an API key, starting on Claude Opus 5.5.
Monthly price of the Nod app and of the ChatGPT plans Nod can sign in with, from n8n’s README and blog and OpenAI’s ChatGPT pricing page. API-key use is billed per token by the provider and is not shown.
Our Numbers
Haiku 5.5’s 8 unsafe actions in 240 decisions are 3.3%, by our arithmetic. Its API output price, $0.50 per million tokens, is also one fortieth of Opus 5.5’s $20, so in n8n’s test the model with no misses cost 40 times more per token, by the same arithmetic.
n8n’s own evaluation reports 98% task completion over 226 errands run three times each, 678 runs with no unsafe action, using GPT-6.1 Sol in simulated accounts and websites. The README calls these “Test results, not a guarantee.”
n8n raised $180 million in a Series C in October 2025 at a $2.5 billion valuation, bringing its total funding to $240 million. Trending Topics reported that a later SAP investment raised the valuation to $5.2 billion.
The Structural Read
As we read the design, the guardrail sits in the server but the judgement inside it is a model call: Nod’s docs say the checker uses your selected model at its lowest reasoning setting by default, so a user who picks a cheaper model for cost also picks it for the check.
As we read the licence choice, Nod ships under Apache 2.0 while n8n’s main product stays under its Sustainable Use License, which limits use to internal business, non-commercial or personal purposes; the agent is the more permissively licensed of the two.
n8n blog, 10 November 2026
“The permission check is itself a model and can make mistakes, which is why Nod asks when your words don’t cover an action”
Three Implications
FOR USERS PICKING A MODEL Claude Haiku 5.5 costs $0.10 per million input tokens against $4 for Opus 5.5 on Anthropic’s pricing page, and let 8 unsafe actions through in 240 test decisions to Opus’s none, per the README.
FOR CHATGPT SUBSCRIBERS Nod can sign in with ChatGPT Plus, listed at $20 a month, or Pro, listed from $100 a month; Pro lists Dot, OpenAI’s always-on agent.
FOR N8N Nod is a separate app that needs no n8n account; the blog says it does not change the direction of n8n, which raised $180 million at a $2.5 billion valuation in October 2025.
The framework says: “The surface, not the model, governs autonomy. The agentic surface — the harness — is becoming where value pools in the entire AI stack.”
As we read it, Nod is a harness given away free on top of any model, and n8n’s own 240-decision test adds a caveat: the surface sets the rules, but the model still decides whether an action fits them. The framework predates Nod and does not discuss it.
What Is Not Established
The 240-decision result is n8n’s own test, published as totals in the README; we did not read the individual decisions and did not repeat the test. The README does not give results for other models, so the comparison covers three models only.
n8n’s blog and README, as we read them, give no download or user figures, and n8n has not said whether Nod will carry a price later. A hosted version is on its roadmap without a date. The SAP valuation comes from a news report, not from n8n.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
n8n has released Nod free under Apache 2.0, with the AI model left to the user. In n8n’s own test, Claude Haiku 5.5 let 8 unsafe actions through in 240 permission decisions while Opus 5.5 and GPT-6.1 Sol let none through, and Haiku’s API price is one fortieth of Opus’s, by our arithmetic.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. We read n8n’s blog post announcing Nod (10 November 2026), the n8n-io/nod README and its GitHub API data at about 10:00 UTC on 11 November, Nod’s website and its docs pages on approvals, evaluation and models, the licence file of n8n’s main repository, OpenAI’s ChatGPT pricing page, Anthropic’s API pricing page and n8n’s Series C post. The SAP valuation is from a Trending Topics report. Percentages and price ratios are our arithmetic. No company was contacted for this piece.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
