Run n8n on Dockhold
Your own n8n, running in minutes, at your own address.
Not a trial and not a shared instance. Twenty dollars a month, no execution limits, no per-run charges, and you can take it elsewhere whenever you want.
Self-host n8n without the self-hosting. You own the instance, and you can move it out whenever you want.
What n8n actually does
It is a workspace where you draw the thing you keep doing by hand, and then it does it. A trigger on the left, steps in the middle, results on the right.
-
It starts on its own
A schedule, an incoming web request, a new email, a message in a chat, a form submission, or a change in an app you use. Something happens, and the workflow runs. You are not in the loop.
-
It joins your tools together
Hundreds of services have a ready-made step, and anything else has an address you can call. If it has an API, it is reachable, and if it does not, it usually has email or a spreadsheet, which are also steps.
-
It decides, loops and waits
Branch on a condition, merge two paths back together, run through items in batches, retry the ones that failed, and wait until something is ready. This is where a toy becomes a thing you rely on.
-
It can think, when you want it to
AI steps with the model you choose and the key you bring: classify, summarise, extract fields from messy text, or run an agent that calls your other steps as tools. Your prompts and your data stay in your instance.
-
It drops to code when the boxes run out
A code step, in JavaScript or Python, for the bit no node covers. You are never stuck at the edge of what the visual editor can express, which is where most automation tools stop.
-
It shows you what happened
Every run is recorded with the data that moved between each step, so a workflow that misbehaved at four in the morning can be read at nine. Fix the step, replay it, move on.
What it plugs into
Hundreds of services have a ready-made step. Anything else has an address, and there is a step for calling one of those too.
-
Talking to people
- Slack
- Telegram
- Discord
- Gmail
- Outlook
-
Where the work is tracked
- Notion
- Airtable
- Google Sheets
- Trello
- Jira
- Linear
-
Selling and supporting
- Stripe
- Shopify
- HubSpot
- Pipedrive
- Zendesk
- Intercom
-
Building things
- GitHub
- GitLab
- Webflow
- WordPress
- AWS
- Google Cloud
-
Models and vector stores
- OpenAI
- Anthropic
- Google Gemini
- Mistral
- Ollama
- Qdrant
-
Everything else
- Any HTTP API
- Webhooks in and out
- Databases
- FTP
- RSS
- Your own code
Names are here to say what n8n connects to. We are not affiliated with any of them, and the credentials for each one live in your instance rather than ours.
Why people want their own copy
All of the above is true of n8n anywhere. This is the part that is only true when the instance is yours.
-
Nothing counts your runs
A workflow that fires every minute costs the same as one that fires every week. There is no quota to watch, no overage, and no moment where a busy day turns into an invoice. This is the single biggest difference and it is why most people start looking.
-
Your data goes where you send it, and nowhere else
Customer records, invoices, message contents and prompts move between your instance and the services you chose. Nobody in the middle is holding a copy for you.
-
As many workflows as you like
The limit is the memory you bought, not a number in a plan. Most people end up with dozens of small ones rather than a few large ones, and that is the cheaper shape here.
-
You choose the version, and when it changes
Your instance runs a pinned version. It does not update itself underneath a workflow that was working yesterday, and you move it when you decide to.
-
You can extend it
Community-built steps and your own code can be baked into your copy of the starting point. That is a self-hosting capability, and it is not available on a hosted plan at all.
-
You can leave
Export every workflow and credential and import them into any n8n anywhere, including one on your own machine. It is the ordinary n8n, not a variant of it.
Start from something
Pick a starting point, or start empty. Either way it is your instance and you can delete whatever you do not want.
-
Blank canvas
An empty n8n, ready for whatever you already had in mind.
You supply: nothing.
Deploy with this -
Daily digest into chat
Collect a few things every morning and post one tidy message to Slack or Telegram.
You supply: A Slack or Telegram bot, which takes about two minutes to create.
not yet run on the pinned version
Deploy blank for now -
Feed watcher
Watch a feed and tell you when something new shows up, without checking it yourself.
You supply: Somewhere to be notified, so a Slack, Telegram or email account.
not yet run on the pinned version
Deploy blank for now -
Form submissions into a table
Catch a form submission and append it to a spreadsheet or a table you already use.
You supply: Access to wherever the rows should land.
not yet run on the pinned version
Deploy blank for now -
An assistant on Telegram
A chat bot that answers with an AI model of your choosing and remembers the conversation.
You supply: A Telegram bot token. A key for whichever AI provider you use.
not yet run on the pinned version
Deploy blank for now -
Inbox triage
Read incoming mail, work out what actually needs you, and flag only that.
You supply: Access to the mailbox you want sorted.
not yet run on the pinned version
Deploy blank for now
The setup we recommend
One answer rather than a page of options. This is what we would pick for someone starting today.
Recommended
1 GB of app power
$20 a month
A gigabyte of app power at $10, plus a gigabyte of dedicated database memory at $10. Your own address with a certificate and the first 10 GB of storage are included.
What it handles
Everyday automations, webhooks arriving at any hour, and a few dozen workflows. Measured headroom under load: measurement pending .
Keep it at one instance. Two copies of n8n both fire every schedule, so your workflows would run twice.
A managed database itself costs nothing: every account has one. The second $10 buys it dedicated memory, which is what our own deploy guide for n8n recommends. We have not yet measured whether the included memory keeps up on its own, so we quote the setup we would actually buy. If the measurement says the cheaper one is fine, this drops to $10 and we will say so.
What it takes today, honestly
We are building the one screen version of this. Until it ships, here is every step, including the three that matter. About measurement pending from the button to a usable editor.
-
Click deploy and sign in
Google or GitHub. The form arrives with the starting point and the name already filled in.
-
Tick the managed database do not skip
Do not skip this one. Without it, everything you build lives somewhere temporary and a restart clears it.
-
Choose 1 GB of app power do not skip
The free allowance is smaller than this app needs, so this is where the payment step happens.
-
Set the encryption key, then restart do not skip
One value that protects every credential you save. The setup page generates it in your browser and walks you through it.
-
Open your n8n and create your account
Your address is ready. The first screen asks you to set an owner email and password.
New versions, on your schedule
Your instance runs a pinned version of n8n and nothing moves it without you. That is the point: a workflow that worked yesterday still works this morning. When you do want the newer one, here is the whole story.
-
Moving up is a redeploy
Redeploy from your app's page and it rebuilds from the starting point, picking up the version that currently pins. Your workflows, credentials and history are in the managed database, so they are still there when it comes back.
-
Export first, every time
n8n changes its own database on the way up, and those changes only run forwards. An export before you upgrade is what turns a bad surprise into a five minute rollback.
-
Want to choose the exact version?
Fork the starting point, change the pinned tag, and deploy from your fork. Now the version is entirely your decision, and the same fork is where you add community-built steps.
What we do not do is update n8n underneath you. We keep the platform it runs on patched. The version is yours, which also means the upgrade is yours to choose, and the same is true of staying put.
What you get that a machine at home does not
-
It stays on
If the app falls over it is restarted and put back. Nothing depends on a machine at home being awake.
-
Its own address, with a certificate
Handled for you, renewed for you. Bring your own domain later if you want one.
-
Its own isolated environment
Your app cannot reach anyone else's and nobody else's can reach yours.
-
A managed database
Attached before the first boot, so there is no moment where your work is sitting somewhere temporary.
-
Hosted in the EU
Along with the rest of Dockhold.
-
Nothing to maintain
The platform underneath stays patched without touching what you built on top of it.
Current platform state is on the status page. We would rather point you at that than print a number.
Where this actually sits
Four ways to run n8n. Each of them is the right answer for somebody, and only one of them is us.
| Option | Better at | Costs you | Right for |
|---|---|---|---|
| A machine at home | Costing nothing extra and keeping everything on hardware you can touch. | It sleeps, it reboots, and nothing on the internet can reach it without work. Every upgrade and every backup is yours. | Learning, and workflows that can wait until the machine is awake. |
| A server you rent and manage | Total control. Any version, any extra software, any layout you like. | You are now the operator. Certificates, upgrades, backups, memory limits and the incident at the wrong moment are all yours. | People who enjoy that part, or who already run other things on the same machine. |
| n8n Cloud | Being run by the people who write n8n, with their support and their own view of what is coming. | Their pricing and their execution limits, which is the reason most people go looking. | People who want the vendor relationship and do not want to think about hosting at all. |
| Here this page | Owning the instance without operating the machine. Your address, your data, no execution limits, and capacity you can split with your other apps. | A card, and a support boundary: we keep it running, and n8n questions go to n8n's own community. | People who wanted to self-host and did not want the maintenance that comes with it. |
Ours is $20 a month at the recommended size. What the others cost, checked on a date we will print next to them: measurement pending . The longer comparison with n8n Cloud.
It is your instance
We keep it online, patched underneath, backed by a managed database, on your own address. The workflows are yours, the data is yours, and you can take it elsewhere any time.
What that means in practice: we do not read your workflows, we do not build them, and we do not debug them. For questions about n8n itself, its own community is excellent and faster than any vendor would be.
If you would rather do it yourself
That is a completely reasonable choice, so we wrote the guide for it. A setup with a real database, real persistence, and honest notes on backups and upgrades. No signup, no gate, and no part of it held back.
It works. What it costs you is the maintenance, forever. If you would rather not, that is what the deploy button is for, and it is $20 a month.
Doing this from an AI assistant
If that is how you work, here is everything an assistant needs. Copy the block, or connect the tools directly.
Deploy n8n on Dockhold for me.
Target: https://github.com/dockhold/n8n-starter
Deploy form: https://app.dockhold.eu/new?repo=https%3A%2F%2Fgithub.com%2Fdockhold%2Fn8n-starter&name=n8n
App size: 1 GB (this app does not run reliably below that)
Database: required, use the managed database, attach it before the first boot
Cost: about $20 a month, a payment method is needed
Instances: exactly one, never more, or every schedule fires twice
Secrets: the platform generates and stores the encryption key. Do not invent one,
do not put one in a URL, and do not ask the user to paste one anywhere.
After deploying, poll the app status until it is running, then give the user the
app URL and tell them to create their owner account on the first screen.Endpoint: https://api.dockhold.eu/mcp
Auth: a Dockhold API token, set in your own tool's configuration.
Never paste it into a web page, including this one.
Scopes: read, deploy
Call in this order:
1. deploy_app repo=https://github.com/dockhold/n8n-starter name=n8n
memory=1024 database=true
2. get_app_status poll until it reports running
3. get_app_logs only if it does not come up
4. resize_app only if the user asks for a different sizeThe full machine-readable version, including how to poll for readiness, is on the page written for assistants. There is also an llms.txt for this site.
The eight questions people ask first
Each one has a fuller answer of its own.
Can I run more than one copy for reliability?
No. Keep it at one. Two copies of n8n both fire every schedule and both answer every trigger, so your workflows run twice and logging in behaves strangely. One instance is the correct setup here.
Running n8n across several copies needs its queue mode, which needs a piece of infrastructure we do not offer today. The size controls let you make one instance bigger, and that is the right lever. If you see an option to add extra instances, leave it at one for this app.
Where did my saved credentials go?
On a self-hosted setup this is almost always the encryption key. n8n encrypts every credential with one key, and if that key is not stored somewhere permanent it is regenerated on restart, leaving credentials that nothing can read.
Here the key is generated once and kept for you, so there is no step to skip and nothing to lose. The full explanation, including how to recover it on your own machine, is on its own page.
My workflows disappeared. What happened?
Something was storing them somewhere temporary. n8n keeps everything in one place, and if that place does not survive a restart then a restart is a factory reset.
Turning on a database afterwards does not bring anything back, it gives you an empty one. Instances deployed from this page get a managed database attached before the first boot, so there is no tick box to miss.
What does it actually cost, and can I cancel?
Twenty dollars a month for the setup we recommend: ten for a gigabyte of app power, ten for a gigabyte of dedicated database memory. Billing is monthly and you can cancel whenever you like. You are buying capacity rather than a plan, so there are no per-run charges and no execution limits to watch.
A managed database is included on every account at no charge. The second ten dollars buys that database dedicated memory, which is query speed rather than the database itself, and it is what Dockhold's own deploy guide for n8n recommends. Capacity is yours to split: buy a gigabyte for n8n, want something else running later, and you divide what you already have rather than buying a second plan.
Will the free allowance run n8n?
No, and we would rather say so here than let you find out. The free allowance is real and it is genuinely useful for smaller apps, but n8n needs more memory than it includes.
That is why every price on this page is a real price and why the word free does not appear next to the deploy button.
Do I need a GitHub account, or to fork anything?
No. Sign in with Google. You are deploying from a public starting point that we maintain, and nothing is copied into an account of yours.
Forking is only worth doing if you want to pin your own version of n8n or bundle extra community nodes into the image. Both are explained on the setup page.
Who can reach my instance?
By default, anyone who knows the address, which is why the first thing n8n asks you to do is create an owner account with a password. Nothing is open before you set that.
Any account that has bought compute, which includes every setup on this page, can also switch the app to private. A private app is reachable only with a token you control, checked at the edge before the request ever reaches n8n. That is a Dockhold setting rather than an n8n one.
How do I update n8n, and do you do it for me?
No, and that is deliberate. Your instance runs a pinned version and it does not change underneath you. You choose when to move, and you do it by redeploying.
Moving up is a redeploy from your app's page. It rebuilds from the starting point and picks up whatever version that currently pins, and your workflows and credentials are in the managed database so they survive it. To control the version exactly, fork the starting point and pin the tag you want. Upgrades change n8n's own database in a direction that does not reverse, so export before you do one. We keep the platform underneath patched either way.