# Nod by n8n: What It Is and What's Known (2026)
By Grace Muthoni Kariuki · 2026-11-11 · Source: https://www.activepieces.com/blog/nod-by-n8n-what-it-is-and-whats-known-2026
---
## What nod is as a local agent
Nod is a local personal agent released by n8n, the workflow-automation maker that sits in the same crowded field as [Activepieces](https://www.activepieces.com); n8n hasn't published a launch date that I could confirm as of 2026-11-11.
Where n8n has published nothing, the cell says so, so treat those fields as open rather than assume them. The card below records what the maker states about nod, with the check date.
| Field | Detail |
|---|---|
| Product | Nod |
| Maker | n8n |
| Launch date | not published |
| Runs where | Locally, on the user's own machine |
| Source page | not published |
| Date checked | 2026-11-11 |
| Activepieces | Maker: Activepieces; launch date: —; runs on Activepieces Cloud or on-prem (self-hosted); 747 integrations and 25,470 GitHub stars, MIT licence on the core; source: https://github.com/activepieces/activepieces |
With no launch date and no source page on record, anyone planning around nod should confirm both with n8n before committing a workflow to it.
10,000 GitHub stars in October 2020 was already a real following, according to [GitStarClub](https://gitstarclub.com/n8n-io/n8n), and it came long before agents were fashionable.
The project reached 50,000 in November 2024, and **100,000 in June 2025**, which means **the second 50,000 arrived in about seven months**.
By September 2026 it stood at 204,100. Stars aren't deployments, but the count shows a large audience that'll test nod publicly.
### What nod is, in n8n's own words
Take "local personal agent" as the only descriptor this article relies on. Local means the agent and its credentials sit on one person's machine, so the first audit question is who owns each connection it holds.
A personal agent isn't a shared team service. Check anything beyond that descriptor against n8n's own page.
## How a local agent benefits teams automating work
That is why a local agent matters to teams: its files and execution stay on one person's machine. The value to a team depends on how a shared workflow can reach it.
### Where the work runs
Start with where each side lives. Nod runs on a laptop and works against the files stored there.
A team workflow runs elsewhere, on a canvas of connected app steps, and each of those connections belongs to whoever authorized it. These are two separate places with two separate owners.
Between the two sides in the illustration below sits one arrow, labelled "call?", a workflow reaching toward the laptop and the one open question between them. The dashed outline marks the data boundary that such a call would cross.
Read the arrow as a question, not a feature. Nothing established so far says a workflow can invoke the agent, and whatever crosses that dashed line is leaving a personal machine for a shared system.
### What it changes for an existing automation stack
Today there's **no native integration to rely on**. A team that wants a workflow to call Nod would be assembling the link from general building blocks, not selecting a ready-made step. That shifts the checklist.
Before wiring anything, settle who owns the connection. If it sits under one person's credentials, the workflow's access to the laptop ends with that person's role, device and availability.
Then settle the credential type and its scope: what the call may ask the agent to do, and which local files it can touch. Only after that does it make sense to weigh the risk.
Apply the same caution to what the maker says the agent can do. Confirm each capability against n8n's own page before a workflow depends on it.
## What's measurably new in nod
There is no confirmed figure for Nod in the record behind this article. The check date is 2026-11-11, and what's measurably new is limited to how n8n describes it and to what n8n's left unpublished.
### Facts n8n states
Because n8n presents Nod as a local personal agent running on one person's machine, **any connection it holds belongs to that person**, not to the team's shared n8n account. Before wiring it into a workflow, ask who owns the credential and what scope it carries.
For a team deciding whether to depend on it, five facts can be checked against n8n's Nod page, and they show which are settled and which stay open.
Platforms, licence and cost are published on Nod's pages, checked 2026-11-11.
Until one of them changes from "not published" to a stated value, treat it as unknown. **Don't hand Nod a token, or build a workflow around it, on the strength of an assumption.**
### What n8n has not published
All of it is open. Without a published trigger or integration, a team can't assume a native node exists, and the practical route is general building blocks such as an HTTP request or a webhook.
Each of those carries its own credential, which someone on your side must own and scope.
When the page doesn't say, a team can't decide who may install Nod or who pays for it, and that is why the missing licence and price lines matter. Recheck the page before each rollout step and record the date you checked it.
## How to connect nod to a workflow in Activepieces today
You connect through its general steps, because Activepieces has no native nod support: its repository and integration catalog, checked on 2026-11-11, contain nothing specific to nod. We haven't run nod, so every route below depends on what nod actually exposes.
Simplest comes first, and each step applies only if the check before it turns up a matching interface:
1. Check nod's docs for an exposed endpoint.
2. Call it with an HTTP request step.
3. If the endpoint is OpenAI-compatible, point an AI step at it.
4. If nod has an MCP server, connect that.
Whichever route you take, settle ownership first: name the credential nod requires, its scope, and the person who'll rotate it.
For the HTTP request route, none of the 747 integrations counted in Activepieces' open source monorepo is specific to nod, so the endpoint nod's docs list is the whole interface, with no integration between you and it.
Note the authentication header nod expects and store it as a connection, not as text typed into the flow.
Because nod runs locally, the machine hosting Activepieces must be able to reach it, which a cloud-hosted instance usually can't. Every Activepieces plan can run on its cloud or on-prem, and the Community Edition is free to self-host under the MIT licence.
That reachability requirement points toward a self-hosted install, where the usual trade is a stripped build with the controls held back.
Activepieces doesn't make that trade: the enterprise controls (single sign-on, SCIM provisioning, custom roles, audit logs, release management and secret managers) are the same list in the air-gapped build as in the managed cloud, so whoever owns nod's credential works under the same roles and audit trail either way.
MoneyGram, Moneypenny, Alan and FundingSocieties run Activepieces in production, and regulated and public-sector organisations run the air-gapped edition today.
If nod's docs say its endpoint follows the OpenAI request format, an AI step can take nod's base URL and key in place of a hosted provider's. Confirm that claim in the docs rather than inferring it, because a near-match fails on the first call.
If nod ships an MCP server, connect it; MCP is the Model Context Protocol, a standard for exposing tools to agents. Record which tools it exposes and what each can change. Read the docs for the transport and authentication it uses before wiring it in.
## How nod compares in cost and speed
Since nod's own price isn't among the vendor-stated figures we confirmed, the comparable costs are the n8n plan that triggers it and the tokens its model provider bills. Both belong to whoever owns the account and the API key, so settle that first.
### What n8n's paid plans cost per month
On n8n's pricing page, Starter is listed at 20 € per month, Pro at 50 € and Business at 667 €.
The jump from Pro to Business is over thirteen times, so a team should confirm the tier it needs before a pilot, not after it.
For model tokens, ([source](https://justinmckelvey.com/blog/gemini-api-pricing)) puts input at 0.75 to 4 $ per million tokens and output at 3.75 to 20 $ per million.
**Output costs five times input** at both ends, so a workflow that asks for long replies costs more than one that sends long prompts.
Read the table as a map of who bills you and where the work runs, not as a ranking. It sets the four options side by side.
Output costs five times input at both ends, so a workflow that asks for long replies costs more than one that sends long prompts.
| Option | Price per vendor page | Where it runs | Setup effort | Source and check date |
|---|---|---|---|---|
| Nod | No vendor price confirmed | On your own machine | No native n8n integration; wire it in with general building blocks | Maker's documentation, 2026-11-11 |
| Hosted AI assistant | No vendor price confirmed | Vendor's cloud | Account and sign-in | None checked |
| Cloud workflow agent | 20 €, 50 € or 667 € per month | n8n's hosted plans | Build inside n8n | n8n, 2026-11-11 |
| Model step in a workflow | 0.75–4 $ input, 3.75–20 $ output per million tokens | Model provider's API | API key in a workflow node | [McKelvey](https://justinmckelvey.com/blog/gemini-api-pricing), 2026-11-11 |
| Activepieces | Free: $0; Plus: $20 a month; Team: $200 a month; Ultimate: custom price through sales | Activepieces Cloud or on-prem | Sign up on Activepieces Cloud; no card needed on the free plan | [Activepieces](https://www.activepieces.com/pricing), 2026-10-23 |
There's no speed claim here, because **no test was run**. Nod's setup is the heaviest of the four because the connection is yours to build and own.
## What to watch next as nod rolls out
Watch five places for a sign that nod has gained a supported route into n8n, and treat anything outside them as rumour.
Until one of them changes, the connection stays yours to build and own, so someone on your team holds the credential it runs under.
### Where n8n would announce nod integration first
If all five stay quiet, keep your general building blocks and don't plan around an integration. The list below shows where n8n would announce a native route first.
- The n8n changelog, for new endpoints
- The n8n docs, for new API or MCP pages
- Licence or pricing changes at n8n
- The [Activepieces](https://www.activepieces.com/pricing) repository (MIT-licensed core, 25,470 GitHub stars), for commits that mention nod
- The Activepieces integration catalog, for an integration named nod
When any of them turns up a hit, who owns the connection changes. A native endpoint or integration would bring its own credential type, so read which scope it asks for before you replace a connection that already works.
If licence or pricing changes at n8n, whether running nod from a shared workflow is covered by your current terms can change with it, and that's a question for whoever signs the contract.
### What to do this week
Name one owner for the current nod connection. Write down the token type it uses, the scope it was granted, and who rotates it. Then put the five pages above into a recurring review with a named reader, and record the date of each check.
When one of the pages changes, confirm the claim against the maker's own documentation before you wire anything. A forum post saying an integration exists isn't the same as a documented endpoint you can hold a credential for.
## Related reading
- [AI Agent Builder for Customer Support Teams](https://www.activepieces.com/blog/ai-agent-builder-for-customer-support-teams)
- [Best AI Agent Builders for Sales Teams, Ranked](https://www.activepieces.com/blog/best-ai-agent-builders-for-sales-teams-ranked)
- [9 n8n AI Agent Alternatives for Marketing Teams](https://www.activepieces.com/blog/n8n-ai-agent-alternatives)
## References
- [GitStarClub](https://gitstarclub.com/n8n-io/n8n)
- [n8n](https://n8n.io/pricing/)
- [Justin McKelvey](https://justinmckelvey.com/blog/gemini-api-pricing)