Lovable vs Nrth Star: what each one actually builds

If you run a marketing agency and you have spent any time near AI tools this year, you have probably been told to try both of these. They get mentioned in the same breath, they are both used by describing what you want in ordinary English, and neither asks you to write code. So the question comes up a lot: which one should we be using?

The honest answer is that they are not the same kind of tool, and picking between them is less about which is better and more about which job you are trying to do.

The short version

Lovable builds software. You describe an app or a website, and it writes and hosts one you can open in a browser. The thing you get back is a product: a page, a portal, a dashboard, a tool someone logs into.

Nrth Star builds automations. You describe a job your agency does over and over, and it builds something that does that job in the background of the tools you already use. The thing you get back is not a page anyone visits. It is work that stops landing on somebody.

Lovable does not build automations, and it is not trying to. If what you want is for unanswered proposals to get chased on day three without anybody remembering, that is not an app, and an app builder is the wrong shape of tool for it.

What Lovable is actually for

Lovable is an AI app builder. You describe what you want, it writes a working web application, and it deploys it somewhere you can send a link. For a certain kind of job it is genuinely very good, and an agency has plenty of those jobs:

• A landing page for a campaign that needs to exist by Thursday.
• A client-facing portal where someone can see the state of their account.
• An internal tool that replaces a spreadsheet three people are fighting over.
• A prototype of a product idea, built well enough to show a room full of people what you mean.

Every one of those has a person at the end of it. Someone opens a URL and looks at something. That is what an app is, and if that is what you need, building it by describing it is a much better afternoon than building it by hand.

What an automation is instead

An automation has no URL and nobody opens it. It sits between the tools your agency already runs on, it waits for something to happen, and then it does the thing a person would otherwise have done.

Take the one nearly every agency has. A deal moves to Proposal Sent in the CRM. Three business days go by. The client has not replied and the deal has not moved. Someone should follow up, in your voice, referencing the actual proposal, and should stop after two attempts.

Nobody visits that. There is no screen. It is a rule that has to be awake at 9am on a Tuesday whether or not anybody is thinking about it.

Can I just build the automation in Lovable?

This is the real question, and it deserves a real answer rather than a no. You can absolutely describe that follow-up rule to an app builder and get code back that does something like it. People do this. It works on the first day.

What you have at that point is an application that contains automation logic, which is a different thing from an automation. The gap between the two is not the code. It is everything around the code.

Something has to wake it up

Application code runs when a request arrives. An automation has to run when nothing has arrived: on a schedule, or on a webhook from a system that changed. That means a scheduler, a queue, and something that keeps running when the browser tab is closed. It is the first thing you have to go and build, and it is not the part anyone was excited about.

Credentials have to live somewhere

To read your CRM, the automation needs access to your CRM. That means tokens, refresh tokens, and the specific misery of an OAuth token that expires at 3am on a Sunday. Somewhere in your generated app there is now a place where a key to your client data is kept, and its security is your problem.

It has to be safe to run twice

The failure mode of an automation is not that it stops. It is that it runs again and sends the same client the same follow-up four times. That needs a record of what has already been done and a check against it before every action, and it needs to survive the job crashing halfway through. This is the least interesting code you will ever write and the most expensive to get wrong, because the person who finds the bug is your client.

Somebody has to be watching

Third-party systems change. A field gets renamed, an API version is retired, a permission is tightened. The automation does not announce this. It just quietly stops doing the follow-ups, and you find out six weeks later when you wonder why the pipeline looks thin.

This is the part that makes the difference between a demo and something an agency can rely on, and it is ongoing rather than a one-off build. An app builder hands you the code and the code is now yours. That is the correct behaviour for an app builder. It is a bad deal for a background job you were hoping to stop thinking about.

You have to know what to build in the first place

A blank prompt is only useful to somebody who already knows what they would type into it. Most agency owners know they are losing time and cannot say precisely where, which is why Nrth Star starts with an audit rather than a text box: it asks how the agency runs, and gives you back a ranked list of where the hours actually go, with a number against each one. You pick from that list. Nothing is connected and no card is needed.

What Nrth Star does with that

Once you have picked something off the audit, the sequence is deliberately slow in the places where being fast is how automation goes wrong.

1. You read the plan before anything exists. Nrth Star writes back what the process is today, what should happen instead, and what it still needs to know. In plain English, not a flowchart. You approve it or correct it in ordinary words.
2. Then you connect an account, and only the ones that plan needs. One click per app, on the sign-in screen you already know. Never before the plan is agreed.
3. Then it is replayed against your last 30 days with everything switched off. No emails are sent and nothing changes. You see what it would have done, against real clients and real dates, before it is allowed to do anything.
4. Then you arm it, which is its own deliberate action. One press turns it off again.
5. Then it is watched. When something breaks, you are told what stopped and offered a fix, and the automation waits for you to approve it. It never repairs itself quietly, because an automation that fixed itself wrong does not fail. It succeeds at doing the wrong thing to real clients.

None of that is a feature an app builder is missing. It is a different product with a different job.

Where the two genuinely overlap

One thing, and it is the reason the comparison keeps coming up: both tools removed the canvas.

The old way to build an automation was a node editor. Boxes, arrows, conditional branches, and an afternoon spent learning what a webhook is before you could describe a rule you could have said out loud in one sentence. Both Lovable and Nrth Star took the position that you should be able to describe the outcome and let the tool work out the middle. That is the shared idea, and it is why they feel similar in the first thirty seconds.

They diverge immediately after, because what gets built is not the same kind of thing.

So which one should you use

Use Lovable when the output is something a person looks at. A site, a portal, an internal tool, a prototype. If you can finish the sentence “and then they open it and see…”, you want an app builder.

Use Nrth Star when the output is that a job stops landing on someone. Chasing quotes, onboarding a client, rebuilding the same report every month, moving information between two systems that will not talk to each other. If you can finish the sentence “and then nobody has to…”, you want an automation.

Plenty of agencies will want both, and they do not conflict. Build the client portal in an app builder. Have the automation be the thing that keeps it fed.

One more difference worth knowing about

Almost every AI builder, Lovable included, prices in credits. You buy a balance, actions draw it down, and working out what a month costs requires a spreadsheet and a guess about how many times you will need to ask for a change.

Nrth Star prices by the number of automations you are running. Not credits, not tasks, not tokens, not seats. Starter is $49 a month for 2 active automations, and the AI writing is included with nothing to buy on top. You know the bill before you get it, which matters more than it sounds when the thing you are buying is meant to be running unattended in the background.

The audit is free permanently, connects to no accounts and needs no card. If you are not sure which of these two tools you need, that is the cheapest way to find out: it will tell you what your agency is actually losing time on, and then it is obvious whether the answer is an app or an automation.