What does this cost?

You are trying to work out what a piece of custom software should cost you, and every source you ask is paid on the answer. This page is the arithmetic instead: what the alternatives charge, with the qualifier each figure came with, and what actually drives the difference between a five-figure quote and a monthly subscription.

What does RunMyB cost?

RunMyB publishes no price on this page. We are not going to invent one, and we are not going to say “free during beta” either — that is a price, and it sets an expectation somebody would later have to break. The platform is in closed beta, by invitation, so nobody is being quoted anything today. When we publish a number it goes on this page, not behind a quote form.

That is an unsatisfying answer if you are trying to put a line in next year’s budget, and it reads badly next to vendors who publish a number, however unpleasant the number is. We would rather be unsatisfying than wrong.

One thing about the shape is already decided and written down. What you pay the platform is its usage pricing: itemised, effective-dated, and shown to you inside the product before anything is charged (Terms §14). We do not publish those numbers on this site.

How much does a custom web app cost to build in 2026?

Agency quotes for a simple custom web app run $15,000–50,000, and mid-complexity work $75,000–250,000 and up on North-American agency figures. A freelancer is cheaper by the hour: on Upwork’s own guide (May 2026), beginners bill $10–25 an hour, intermediate freelancers $25–60, and advanced or specialist developers $60–120 and up. Neither figure covers what happens after launch.

Where you hire changes the hourly rate more than what you are building does: $80–140 in North America, $75–95 in the UK, $70–85 in Germany, $40–70 in Eastern Europe and $30–55 in Latin America. Senior Polish developers run $60–90 an hour. Agency rates follow the same geography: $150–250 and up an hour in San Francisco, $100–175 in Canada, $30–80 for Eastern-European and South-Asian teams.

If you describe the same idea to three people and get three numbers an order of magnitude apart, that is not dishonesty. It is that nobody has written the specification yet, so each of them is pricing their own guess about how bad it gets. The scoping phase that would settle it is itself something you pay for, and it is quoted by the people who only earn if you proceed.

There is a second bill nobody puts in the quote. A freelancer hands you working code and you now own the hosting, the payment wiring, the sign-in system and every upgrade. That work does not disappear because the invoice ended.

Here there is no specification to write and no scoping phase to pay for. You describe the job in conversation, the build starts from that, and you drive the working product on a private preview before anything is published. Compared with the alternatives on this page — an agency, a freelancer, a no-code subscription, a first developer hire — that is a different order of cost. It is also a different order of certainty, which is the honest half of the sentence.

How much do developers charge to change one page?

At whatever their hourly rate is: $25–60 an hour for an intermediate freelancer and $60–120 and up for an advanced one on Upwork’s own guide, $80–140 in North America, $75–95 in the UK. Where an agency is retained at all, upkeep runs 15–20% of the build cost a year and small edits queue inside it. The invoice is rarely the real cost. The real cost is that you learn to stop asking, and a product that stops changing stops being right about your business.

It is your product, one number on one page is wrong, and you are waiting two weeks for somebody else’s next window. Six months after launch you are asking permission to edit your own business.

Changing a product you own here never re-enters a paid queue. You describe the change in plain language; the platform makes it, re-certifies the whole build against the same floor it had to pass the first time, and stops at an approval gate. Nothing ships until you approve it. No invoice, no two-week window, no negotiation.

What does it cost to keep a custom app running after launch?

Post-launch upkeep is priced at 15–20% of the build cost a year where anyone is retained at all, and on most builds nobody is. The code is delivered, the team moves on, and the first real fault arrives two years later as a quote larger than the original build: the framework is several versions behind, the person who wrote it is gone, and somebody has to read all of it before they can safely change any of it.

That is the line people leave out of the comparison, and it is the one that runs forever. A five-figure build with nobody operating it afterwards is not a finished purchase. It is a deferred one.

There is no handover here, because nothing is handed over. The same system that built the product re-certifies it in place and fixes it, so keeping it working is the ordinary way of working rather than a fresh quote every time something needs attention. Each product also keeps an automatically maintained record of what was decided and why, so the context does not leave when a person does. And the code is yours: it exports to your own GitHub or GitLab (Terms §12.1) — as a fresh single-commit history, so the build history itself does not travel.

Is it cheaper to hire a developer?

A first technical hire runs roughly $100,000–150,000 fully loaded in the first year — salary plus benefits, tax and equipment. Below that, an intermediate freelancer on Upwork bills $25–60 an hour and a senior Polish developer $60–90. The difference is not only price: a hire buys ongoing capacity and a freelancer buys a finite number of hours. For a product earning four figures, neither side of that trade works.

The advice everyone gives at this point is to find a technical co-founder instead, which is the same bill paid in equity.

What a platform like this removes is the reason to buy developer hours, not the reason to have judgment. Describing a change and getting it built and re-certified is the ordinary way of working here, not a quoted project. It does not remove the person. Somebody still has to talk to customers, decide what to build next and approve what ships, and the platform is built to stop and ask at exactly those points. We are not going to tell you to fire your developers.

One number we will not give you: the fully loaded cost of a first operations, admin or marketing hire. The $100,000–150,000 figure is a software developer, and re-labelling it would be the easiest sentence on this page to write and the least true.

What does a no-code or creator-platform stack cost per month?

The published tiers, each with the qualifier it came with: Bubble $29, $119 or $349 a month for a web app billed annually ($32, $134 or $399 billed monthly), plus metered workload units; Wix $17–159 a month on a yearly plan paid up front; Kajabi $143–399 billed annually ($179–499 monthly), taking no revenue share but 2.9% + $0.30 card processing on Basic and a further 2% / 1% / 0.5% by tier if you bring your own payment provider; Teachable $29–139 billed annually ($39–189 monthly) with 7.5% on the entry Starter plan and none above it; Podia $42–150 billed annually ($49–179 monthly) with 5% on the entry Mover plan; Gumroad no monthly bill but 10% + $0.50 on a sale through your own links and 30% through its Discover marketplace; Whop no monthly fee and no marketplace commission, at 2.7% + $0.30 per card sale.

Read that list twice and the structure appears. Every one of those bills buys hosting and selling. Not one of them builds software. So the calculator, the generator or the client tool your customers keep asking for arrives as an additional capital line on top of a stack you are already paying for — and the stack grows while the thing you wanted never appears.

The second structure is the cut. Move platform to escape it and it changes shape rather than disappearing: a percentage of every sale on the entry tiers, a monthly subscription instead of a percentage further up. What differs is what the bill buys: on a storefront it buys a checkout wrapped around a file you already made; here it buys the product itself, built for you, hosted, and still changing.

If a percentage of every sale is what you are optimising, optimise for it — and here is the whole rate on each, because half a rate is worse than none. Gumroad has no monthly bill and takes 10% + $0.50 of a sale made through your own links, but 30% when the buyer finds you through its Discover marketplace. Whop has no monthly fee and no marketplace commission, at 2.7% + $0.30 per card sale, more for international cards or currency conversion, and some of its add-on apps carry their own percentage. Also be clear about the trade in the other direction — building the tool your stack has never had retires one line item, not four.

Why do AI builder credits run out faster as the project grows?

Because the meter is weighted by work done rather than by requests made, and the work behind one change grows with the project. The bigger the codebase, the more has to be read before anything can be written, so month three costs more per change than month one did. Bubble bills metered workload units on top of $29, $119 or $349 a month for a web app billed annually. The ceiling is usually discovered after the spend, not before it.

The other half of the same problem is that the tools are built around a session rather than a project. Nothing persists a plan, a decision or a reason between sittings, so you re-explain your own product to it every time, and you pay for the re-explaining.

Here a long project is run like a project. Every product carries an interlinked record of what was decided and why, grown only from real sources, so plans and reasons survive between sittings. Work checks the balance before it starts, so a run that cannot be paid for never begins rather than stopping halfway. Spend on the metered AI door is capped per grant. When a ceiling is reached the work pauses instead of dying, so a half-finished product waits for you rather than being abandoned mid-change.

How much does it cost to host a small app every month?

We do not publish a RunMyB number for that yet, so this answer is the shape rather than the figure — and the shape is the part that catches people. Every hosted product here gets its own database and its own sandbox — that is what keeps one product’s trouble away from another product’s data — and it is a real recurring cost per product, whether or not anyone is using it. (that shape is the same whether the product is a client-side tool, one with its own database, or one with its own end-user accounts; we review a product before it is served publicly)

That matters most if you took the standard advice and launched four small things to see which one worked. Three flop and they still cost you every month. A storefront hides this shape, because a dead product on a percentage-of-sales platform costs nothing to leave up. Per-product hosting does not work that way, and you should know that before the fourth launch rather than after it.

One thing follows, and it is the practical part: you can suspend public serving on a product and resume it later without re-applying and without deleting anything, so a quiet product does not have to be a destroyed one.

Will this cut our SaaS bill?

Probably not the way you are hoping. The waste is real and measured: 73% of SaaS vendors raised their prices by an average of 12% between 2022 and 2023, and nearly 50% of SaaS licences go unused for 90 days or more. RunMyB does not find that waste. There is no licence discovery here, no spend analytics, no SaaS management, and none is planned.

What it does instead is replace specific line items: a custom internal application, or a point tool somebody bought for one job, rebuilt as the thing you actually need and run for you. That is line-item substitution, not stack replacement. Your accounting system is not on the list.

Anyone selling you “cut your software bill with AI” is selling a different product. Make them show you the discovery mechanism before you believe the saving.

What does it cost to leave?

Nothing to press the button. The cost is what you have to stand up on the other side. Export pushes a product’s current code into your own GitHub or GitLab account, under your ownership (Terms §12.1) — as a fresh single-commit history, so the code travels and the build history does not. Hosting and the database are things you would set up elsewhere, and we would rather say that now than let you find it on the way out.

A full-tree secret scan runs before every push and refuses the push when it finds likely credentials — a safeguard operated on a best-effort basis, not a warranty that every secret is caught (Terms §12.4). From there you can keep the product in two-way sync, or bring a repository you already own and have it managed onward.

Compare that with what you are leaving today. A course platform, a link-in-bio store or a community exports your content, never the thing that runs it. A no-code application at $29, $119 or $349 a month for a web app billed annually, plus metered workload units, is not separable from the platform at all. What you take out of those is a screenshot and a CSV.

Test your export early, while nothing is wrong, rather than on the day you are angry.

What we cannot put a number on

Some of the costs you are about to meet are real, and we hold no figure we can honestly quote for any of them. Every number on this page came from a survey with a source behind it; these did not, so they get named without one. A cost you were warned about and have to price yourself is still better information than a plausible figure somebody made up for a web page. Here they are.

Where to go next

The home page covers what the platform builds and what it deliberately does not. Access explains how the closed beta works and what happens if you are invited. The commitments cited above — ownership and export — are published in full in the Terms of Service & Acceptable Use, and you can read them before you talk to anyone. To ask a question, write to hello@runmyb.com.