In February 2026, two weeks after the first commit, Kata Jaam went live on both app stores. It is an event-discovery app for Nepal that I founded, and I built it on a stack chosen for being boring. It had no custom backend server, no native plugins and no payments. It was my own product, so there is no invoice to quote.
Flutter app development cost, for most of the products I am asked to build, lands between USD 8k and 30k for a first store release. I charge USD 60 to 90 an hour, or a fixed price per phase, as of October 2026. Those are one engineer's bands, built from thirteen apps shipped to the stores between 2021 and 17 Oct 2026, not a market survey. Kata Jaam is where the low end comes from. Most of what follows is about what pushed the other projects toward either end.
Two weeks and a boring stack
The cheapest version of an app is the one that does one job well. Kata Jaam's job was to tell you what is on in Kathmandu tonight. The stack was Flutter with Riverpod, Firebase for data and Cloud Functions, Next.js on Vercel for the web, PostHog for analytics and Cloudinary for images. Nothing in it needed Swift or Kotlin.
A client build of the same scope would sit near the bottom of the band above. A lean v1 still needs store listings, crash reporting and a release pipeline, so the floor is not zero. It stays cheap because I say no to features until the first users ask for them.
What pushed the other apps up the band
In my experience, five things move a quote the most. Livio, a subscription app for people losing weight in the United States, hit three of them in one project.
The first is native work. Health data, calls, payments and the camera all need Swift and Kotlin beside the Dart. For Livio I built Apple Health and Health Connect sync, a camera-based vital-signs scan and subscriptions on StoreKit and Play Billing. Calilio had the same problem in a different shape: its calling engine was a custom Flutter plugin over Twilio's native voice SDKs. Each of these is days or weeks of work most Flutter developers avoid.
The second is backend scope. Firebase was enough for Kata Jaam, and it is enough for many v1s. Livio needed its own Python and PostgreSQL backend, plus an agent that calls 30+ tools, which made it a different project.
The third is an existing v1 to rebuild. Its version 2 replaced a product members were already using. Rebuilding means reading what exists, keeping what users depend on and migrating data. It usually costs more than starting fresh would have.
The fourth can move a project down as easily as up: two stores or one. One Accord: Scanner shipped on Android only. Two stores means two review processes, two billing systems and two sets of device quirks, even with one codebase.
The fifth is design. If you bring finished designs, I build from them. If you do not, someone still has to design every screen before it is built, and that time goes into the quote whether I do it or a designer does.
Store submission, the release pipeline and over-the-air patching are on the deliverables list for every build I quote, along with a written handbook.
What kept costing after the stores said yes
Store accounts are paid by you, in your name. Apple charges an annual developer fee and Google charges a one-time fee to open a Play Console account.
Backend bills depend on how the code talks to the database, not only on how many users you have. Kata Jaam taught me that one after launch. The Firestore Admin SDK's default gRPC transport wedged on reused Vercel instances and turned a cache into about three million Firestore reads a day. Switching the SDK to REST, one flag, cut the bill to a flat rate. The whole hunt is written up here.
Crash reporting and analytics are the next line. I wire Sentry or PostHog from day one. On Baideshik Rojgar we learned about slow screens from reviews before we learned from data, and I would add monitoring earlier if I did it again. Both tools have free tiers that cover a small app for a while.
Then releases. Baideshik Rojgar passed 1,000,000 installs on Google Play at a 4.4 rating while I led its mobile side, from October 2023 to June 2024. At that scale every rough edge is a thousand support messages. The discipline I use now started there: staged rollouts, automated release checks, a verification script before every upload. Over-the-air patches with Shorebird mean a Dart fix can reach users without waiting on store review. None of that is free to maintain, so budget for work after launch the way you budget for hosting.
Questions to ask any Flutter developer before you sign
Whoever you hire, me or someone else, I would ask these before signing.
- Will I see a demo build on my phone every week? Status reports are easy to write. A build is hard to fake.
- Do I own the code and the store accounts from the first commit? If the developer holds them, you are renting your own app.
- Is there a written handbook at the end: how it is built, what not to break, how to deploy?
- How are AI coding agents used, and who reviews their work? I use them to do the typing, under my review. That is how one senior engineer can quote a one-person price.
- Which parts need native code, and have you shipped that before?
The last one matters most. A developer who has never built a CallKit integration or a subscription flow will find out how long it takes on your budget.
Where my numbers stop
My bands come from my own projects: agency builds for US and Australian clients, two US restaurant chains' ordering apps, a political party's app in Nepal, a job board for migrant workers, a health coach, my own product. They are not a survey of what agencies charge. A team of five in London or San Francisco will quote differently, and some projects fall outside the band in both directions. The reason is usually scope. A small single-purpose app comes in under it, and heavy native work on top of a custom backend goes over.
I have not built a game, an AR or VR app, or a React Native app in production. If your app is one of those, my band does not apply to it.
If you want a band for yours
The build service has the phases, the rate and an assistant that gives a rough band for your app.