n8n, the German workflow-automation company, has put a new product on Product Hunt called Nod: an open-source personal agent that runs on its own computer rather than inside a chat pane. The listing is thin on specifics, which is itself the story. What n8n is selling is not a better chatbot. It is the claim that an agent needs a machine, and that the machine should be yours.
That framing matters because it cuts against where most of the money has gone. The dominant agent products of the past two years have been cloud-hosted and vendor-controlled. You describe a task, a model plans it, and the vendor’s servers execute the tool calls. Nod inverts the arrangement: the agent gets a computer, presumably a sandboxed environment, and the open-source license means you can inspect and run it yourself. n8n has spent years building exactly the connective tissue this requires. Its core product is a node-based workflow builder that wires APIs, databases, and SaaS tools together, and it has an established open-source community plus a hosted tier. Nod looks like that same wiring, pointed at a single user instead of a company’s ops team.
Why “its own computer” is the load-bearing phrase
An agent with a computer can do things a chat agent cannot. It can install software, keep files between sessions, run a browser, and recover from its own mistakes by looking at the screen. Those are the capabilities that separate a demo from a coworker. They are also the capabilities that make security teams nervous, because an agent with a shell is an agent that can be prompt-injected into deleting things.
The open-source angle is n8n’s answer to that fear, at least in marketing terms. If the agent runs on your box under a license you can read, you can audit what it touches and constrain the blast radius. That is a real argument, and it is the same one that has driven adoption of local model runners and self-hosted automation. It is also incomplete. Open source tells you what the code can do. It does not tell you what a model will decide to do when a malicious web page tells it to do something else.
The economics n8n is quietly betting on
Every agent that runs on its own computer is a computer someone has to pay for. That is the business logic under the open-source veneer. n8n’s hosted product already sells compute and orchestration to teams. A personal agent that wants persistent storage, a browser, and a runtime is a natural upsell: free to self-host, paid to have someone else keep the lights on. The Product Hunt listing does not name a price, and Tessera could not confirm the hosted tier’s terms from the listing alone. {/* TODO: verify Nod pricing and hosted-tier details — searched the Product Hunt listing, no pricing surfaced */}
The compute question is where this gets interesting for the AI economy. If personal agents with their own machines become common, the demand curve shifts from inference tokens toward always-on sandboxes. That is a different kind of spend, closer to renting a small VM than to buying API calls. It favors vendors who already sell persistent infrastructure, and it squeezes the pure-inference players who have optimized for cheap tokens. n8n, with a workflow engine and a hosting business already in place, is positioned for that shift in a way a model lab is not.
The part the listing does not answer
“Nod by n8n” is a name and a one-line description. The source material gives us an open-source personal agent with its own computer and nothing about which models it runs, how it sandboxes execution, or what happens when the agent goes off the rails. Those are not footnotes. They are the whole product.
An agent is only as good as the model driving it, and n8n is not a model company. The sensible read is that Nod is model-agnostic, wiring whatever API or local model the user supplies into the node graph. If so, n8n is competing on orchestration and trust rather than on intelligence, which is a defensible place to stand. The labs keep shipping better models; somebody has to give those models hands and a desk. {/* TODO: verify which models Nod supports and whether it runs local models — not stated in the listing */}
The trust problem is the harder one. A personal agent with its own computer is a personal agent with your files, your accounts, and your credentials. The failure mode is not a bad answer. It is a bad action taken confidently. Open source helps here, but only for users who can read the code and configure the sandbox. Everyone else is trusting n8n’s defaults.
The agent that gets its own computer also gets your keys. n8n is betting the open-source license is enough to make that trade feel safe.
What this means for builders
For anyone building agents, Nod is a signal about where the tooling is heading. The chat-wrapper era is closing. The interesting work is in giving agents persistent environments, guardrails around what they can touch, and a clear story about who owns the machine when something goes wrong. n8n is not the first to say this. The local-first and self-hosted crowd has argued it for a while. What is new is a company with an existing automation business and a real community putting its name behind it.
The open question is whether personal users will pay for a computer to run an agent they could partly get from a cloud product. The answer probably depends on how much they care about control, and how much of their life they are willing to hand to a process that acts on its own. n8n is betting some of them care a lot.
Watch the sandboxing details when they surface. That is where Nod either earns the trust its license implies or becomes another agent with a shell and a cheerful README.