n8n launches Nod, a personal AI agent with its own browser

n8nNovember 11, 2026Watch the video
Nod is a new AI agent that can carry out tasks on your computer, work in your apps, and follow up without you having to manage every step. It is free and open source, and you can choose which AI model does the work.
n8n Labs, an experimental development team at the company behind the n8n automation platform, launched Nod on November 10, 2026. It is currently available as a research preview, an early test version that is still in development.
From answering questions to getting work done
A typical AI chatbot can help you write an email, analyze information, or explain how to tackle a task. Nod is designed to go further and do the work.
In n8n’s demo, Nod prepares for a client meeting by reading past emails, looking at the client’s website, and finding relevant news. It then writes a short meeting brief. After the meeting, it drafts a follow-up email and spots a question from the client that went unanswered in Slack.
These examples come from n8n’s own demo. The documentation describes the features that make this kind of work possible.
The same approach can apply to other tasks. In purchasing, Nod could visit supplier portals, retrieve invoices, check prices, and follow up on orders. In logistics, it could collect information about deliveries and exceptions. Customer service, finance, and marketing are other areas where it could help with tasks that require information from several systems.
You describe what needs to get done, and the agent works out which steps will get you there.
Nod has its own computer and browser
Nod has its own isolated computing environment, often called a sandbox.
The agent works in a restricted Linux environment with its own files, tools, and Chromium browser. Its access to the rest of your computer is limited.
Nod can connect directly to services such as Gmail, Google Calendar, Slack, and Notion. It uses integrations that let different programs communicate with one another.
It can also work on websites that have no such integration. It uses the browser to open pages, fill out forms, navigate menus, and retrieve information much as a person would.
That can help with older systems, supplier portals, and other services where people normally have to sign in and do the work by hand.
Nod uses a server to coordinate tasks, separate environments for files and browsing, and a database to store information. The standard desktop app installation handles that setup for you.
An AI agent with memory
Nod has a memory system that lets it retain information across conversations.
For example, you can tell it who now handles invoices or how you prefer to word your emails. The agent can use that information in later tasks.
You can see and edit the memory. Nod shows what it saved and where it came from, and you can change or remove it.
The developers made a deliberate distinction here. Nod learns personal information and preferences from what you tell it directly. One exception is that it can review your own sent emails to learn your writing style. Other emails and websites are treated as outside information.
Nod can also learn procedures. If it finds a reliable way to retrieve an invoice from a portal, it can save the steps and try them again later.
Remembering who you are and learning how to do a job serve different purposes.
Nod can keep working while you are away
Nod supports scheduled tasks that run at specific times or on a recurring schedule.
You can ask it to check a supplier’s prices every morning and notify you when something changes. It can also prepare a daily meeting overview, watch for new emails, or check whether a task needs follow-up.
The agent can handle several tasks at once. You can talk to it in the Nod app, on Telegram, or on WhatsApp.
It can also suggest tasks based on activity in connected services if you turn that feature on.
The computer running Nod has to stay on and awake. If your laptop goes to sleep, scheduled jobs stop too. You can run Nod on a server or a computer that stays on.
How does Nod keep the agent within its permissions?
Giving an AI agent access to email, documents, websites, and payments makes security essential.
Nod uses a control system called a trust check to evaluate proposed actions against what you actually asked it to do.
If you ask Nod to move a meeting and notify the attendee, it can take the necessary steps without asking for permission at every turn. If it wants to send information to another recipient or do something outside the instruction, it is supposed to ask for approval.
You can also set your own rules. For example, Nod can be required to ask before emailing anyone outside your company.
Passwords and payment information are kept separately in encrypted storage called the Vault. When Nod signs in to a website, the system fills in the credentials without revealing the password to the AI model.
Nod treats content from websites, documents, and emails as information. That helps reduce the risk of prompt injection, where someone tries to steer an AI agent through text it reads.
The system also relies on AI models, which can make mistakes. The developers say these safeguards cannot catch every error or attack.
Does running Nod locally keep all data on your computer?
Some data stays on the computer, and some can go to other services.
Nod stores files, memory, and sign-in information on the computer or server where it runs. When you use a model from a provider such as OpenAI or Anthropic, relevant information has to be sent to that provider for processing.
That can include conversations, document contents, and screenshots from the tasks the agent performs.
Connected apps also typically use a service from n8n and Composio. External services process app requests and results even while Nod itself runs locally.
According to the privacy documentation, sharing basic usage data with n8n is on by default and can be turned off. Sharing conversation traces for debugging is off by default.
Where the program runs, where information is stored, and which services process that information are separate questions.
This matters particularly for organizations that handle sensitive customer information, health information, or other data subject to strict requirements.
You can choose the AI model
Nod is open source under the Apache 2.0 license and is free to download and use.
It supports models from OpenAI, Anthropic, Google, and other providers. You can use an API key, and the documentation also describes signing in with ChatGPT Plus or Pro.
You can also run a model locally through a tool such as Ollama or LM Studio. That keeps the model requests themselves on your machine.
Local models need to be capable enough to handle tool calls and large amounts of information. A compatible model may still be less reliable than the most capable cloud models.
If you use a paid API, the model provider charges you. Nod shows usage and costs per task and lets you set monthly spending limits for API use.
Nod can be installed on Macs running macOS 14 or later, Windows 11, and Linux. Windows and Linux support are still at an early stage. You can also run Nod on your own server.
How well does Nod work in practice?
n8n has published its own evaluations of how well the agent completes tasks, respects permissions, and remembers information.
Nod completed 98 percent of tasks in a test of 226 distinct tasks, each run three times. Across 678 runs, the developers recorded no actions that exceeded the simulated user’s permissions.
In a separate memory test, Nod answered 93.2 percent of 500 questions about earlier conversations correctly. Another test tried to plant false memories through email 216 times. None of those attempts succeeded.
The test conditions matter when interpreting these results.
The developers ran the tests themselves, primarily with GPT-6.1 Sol, in simulated apps and websites. They made the test tools open source. The results describe performance in those simulations.
Nod is still a research preview. It can make mistakes, and its features and reliability may change as development continues.
How is Nod different from n8n?
Nod and n8n come from the same company and serve different purposes.
n8n is primarily a platform for building automated workflows. A workflow might start when an order comes in, update a customer system, send an order confirmation, and create an invoice.
Nod is a personal AI agent. You describe the task in plain language, and it plans and carries out the steps using the tools and services it can access.
You can use Nod without an n8n account.
The two can also work together. According to the documentation, Nod can connect to n8n and help you build, test, and fix workflows. New workflows can stay in draft until you choose to publish them.
That lets Nod do work itself and help you set up more traditional automations.
Key takeaways
Nod shows how AI agents are moving from tools we ask for help to digital coworkers that can carry out and follow up on tasks across systems.
It brings together its own browser, access to apps, memory, scheduled tasks, and controls on what the agent can do.
Nod is still in development. You have to decide which systems, information, and tasks it makes sense to give it access to.
The difference from a typical chatbot is clear. Nod is built to do the job you give it, beyond explaining how to do it.