n8n on your host or A system in our care
n8n after hours.
Name a reader.
A run dies after the office closes, and payroll includes nobody who can read the source. n8n addresses teams that are technical. They may work visually. They may work in code. That code may be JavaScript, and it may be Python. Deployment can sit on machines the company operates, or on machines n8n operates. GitHub is where the source is kept. A hosted version is there for people who do not want the box. Integrations in the catalog number 500+. Human-in-the-loop approvals are included, so a run is able to wait. An n8n consultant can raise the first workflow. Keep the product when a technical colleague will take hosting and can read a workflow once a run fails. Ask us to operate the system when that colleague is missing.
Quick verdict
Operate n8n yourselves when a technical colleague already employed will host it, on machines you run or on n8n, and can read the workflow once a run fails. Let us operate the system when the job should be done without hiring that colleague into the stack. Figures for money, times you would promise, markdowns, a bigger scope, and claims not made before all wait on a person.
Side by side
n8n on your host vs A system in our care at a glance.
A workflow a technical colleague is able to host and change, set next to one we assemble and operate without assigning that colleague the job.
| Dimension | n8n on your host | A system in our care |
|---|---|---|
| How you build | Visually, or in code. JavaScript is one option for that code. Python is the other. The audience n8n names is technical. | The workflow is built by us. Your technical people do not have to inhabit the editor for the job to continue. |
| Where it runs | Infrastructure the company operates, or infrastructure n8n operates. Hosting is offered as a version you do not box yourself. The source is kept on GitHub. | Running what we build is our side of the work. A host, a deploy, and a repository are not choices you have to staff. |
| Approvals | Because human-in-the-loop approvals are in the product, a person can hold a run before the next step goes on. | Anything a customer could rely on is approved by a person you choose. Those messages do not leave while nobody is looking. |
| Integrations | 500+ integrations are what the catalog on n8n.io lists. Check the ones this job depends on before you trust them. | Wiring follows the systems this job touches. Catalog length matters for those apps and not for the rest. |
| A failed run | Reading happens on the canvas, or inside the code, by a technical person, and that person fixes the break. | Reading and repair are ours. The explanation comes to you. A repository to debug does not. |
| Keep this path when | A technical colleague already wants the source, already wants a voice in the host, and already reads failed runs. | The job is plain, and the alternative would be a hire whose main duty is watching the workflow tool. |
Where it may be hosted, where the source lives, how many integrations are listed, and how human-in-the-loop approvals work are written up on n8n.io. Read that write-up before a workflow holds a customer record.
Choose n8n on your host
When this path fits.
- Someone technical and already employed wants the source and will read it after the building is quiet.
- The host preference is already known: company machines, n8n machines, or the hosted version.
- For the maintainers, JavaScript inside a workflow is welcome, and so is Python.
- Pausing a run for a person already matches how you work, which is what human-in-the-loop approvals are for.
- An n8n consultant can compress the opening, after which your own person takes the keyboard.
Choose A system in our care
When this path fits.
- Holding n8n would require a hire who is not here today.
- You do not want GitHub, deploys, or a hosting argument attached to this particular job.
- The workflow should run, and the people who read a failed run should be us.
- Output might be seen by a customer, so approval happens before it leaves.
- What an n8n consultant would leave is a workflow the current staff could not explain a week on.
How we would actually decide
Hold n8n when a technical colleague will read source.
Work a team still does by hand is what we wrap in a system, with AI on the steps that repeat, and a person you trust keeps the commitments: price, time, discount, scope, and any claim that is new. Among the do-it-yourself tools in this set, n8n is the strongest when the team can read code or a canvas, whether that code is JavaScript or Python, and whether the machines are yours or n8n's. The source is on GitHub, a hosted version is there when the box is unwelcome, the catalog lists 500+ integrations, and a pause for a human is what n8n calls human-in-the-loop approvals. Consultants who know n8n can make the first workflow shorter. The reader of a failed run is still a separate requirement.
Run n8n if that reader is already employed and wants the keyboard. The product exists so a team that likes source and hosting can have both. We will tell you that. Sending every company to an outsider would throw away a capable tool.
Stop if the technical reader would be a role you still have to open. Read the hire comparison, or ask us to operate the system. Failures are read by us. What a customer receives still needs a yes from someone on your side. The workflow, and the reader's name, are where Get your free Systems Plan begins. After that name is ours, Follow the build once it starts is the loop.
Frequently asked
n8n on your host vs A system in our care questions answered.
Which kind of team is n8n built for?
n8n is built for teams that are technical, and it lets them work visually or write code in JavaScript or in Python. They can place it on infrastructure they run or on infrastructure n8n runs, and a hosted version is available when they do not want the machine. GitHub is how they read the source, the catalog reaches 500+ integrations, and a human can hold a run because human-in-the-loop approvals are built in. When the team matches that description and someone will read a failed run, n8n is a serious way to keep the work in house.
Is an n8n consultant the right long-term plan?
As an opening, yes. As the ongoing plan, only if a technical colleague already exists. An n8n consultant can put the first workflow together, choose a host, and demonstrate how a run gets read. Without that colleague, the engagement ends, the repository remains, and the reader was never hired. Either ask us to operate the system, or bring that technical hire on and hand them n8n in earnest. An invoice does not create the reader the product expects you to have.
Should the workflow run on our machines or on the hosted version?
n8n supports more than one shape of hosting. You can run it on infrastructure you operate, on infrastructure n8n operates, or on the hosted version. Take your own machines when a technical colleague wants the machine, wants the GitHub copy of the source, and already patches this kind of thing. Take the hosted version when that person wants the workflow and does not want the machine. A reader for failed runs is still required, and neither the host nor the repository provides one.
How do human-in-the-loop approvals fit either choice?
They are included, which means a run is able to pause until a person continues it. Host the workflow yourselves and use that pause when you want a halt partway through. On a system we operate, the parallel is a person agreeing to what a customer is about to get, before sending, covering a price, a promised time, a discount, a wider scope, or a fresh claim. Who presses the approval matters more than which product draws the pause.
Can we hire a technical person after the system is already running?
You can. One order is that we operate the system, and a technical hire comes after you understand the workflow and the volume has leveled off. The other order is the hire first, with n8n placed in that person's hands immediately, source included and host included. Each order is coherent. Putting n8n in place, never hiring, and expecting a non-reader to debug a failure because the diagram looks friendly is not coherent. Friendliness in the diagram is not a person.
What should we record before we choose a path?
Three notes. The workflow itself. Which host you can accept. Which technical colleague would read a run that failed. A name already on staff makes n8n reasonable, and a consultant stays limited to the opening pass. A name that would be a new seat means you choose between opening the seat and asking us to operate the system. Get your free Systems Plan is those notes set against the job you have, rather than a walk through features.
More comparisons
Other decisions worth getting right.
Next step
Record who can explain a run that failed.
Someone already employed who wants the source makes n8n a serious route. An empty line means the write-up we give you is the system we would operate.