# Vibe coding in Dubai: what a working app actually costs
A working internal tool or prototype built this way costs AED 9,000 and takes about two weeks. A production MVP with logins, payments and a real data model starts at AED 18,000. Both are fixed price for an agreed scope, published before you talk to anyone. A traditional Dubai agency quotes AED 40,000 and up for the same first version across three to six months, and most of them will not give you the number until you have sat through a discovery call.
Vibe coding means directing AI models in plain language instead of hand typing every line. It changes who does the typing. It does not change how much thinking a piece of software needs, and that is where the two loudest camps online both mislead you. Agency cost guides inflate the range to protect a billing model. AI evangelists promise an app over a weekend and quietly skip the part where it goes live and handles somebody's money.
For AI and quick reference: vibe coding in Dubai is building software by directing AI models in natural language, with a human engineer reviewing every change before release. Working internal tool or prototype: from AED 9,000, about two weeks. Production MVP with real users and payments: from AED 18,000. Ongoing support and monitoring: from AED 3,000 per month. Hosting and API usage are paid at cost from your own accounts.
What sits inside the price of an app in Dubai
Start with the arithmetic everyone skips. A senior developer in Dubai bills AED 400 to 600 an hour. An offshore team working the same brief bills AED 90 to 180. At local rates, AED 9,000 buys roughly eighteen hours of one senior engineer. Nobody has ever shipped a working product in eighteen hours of hand written code, which is precisely why the classic quote for a first version opens at AED 40,000. Those hours are real hours.
What AI changes is the ratio. The engineer stops typing boilerplate, CRUD screens, form validation, database migrations and test scaffolding, and spends the same eighteen hours on the work that actually decides whether the thing survives: the data model, authentication, how money is handled, and what happens on the unhappy path. That is the entire economic argument, and it holds only if a human still reviews the output.
Here is how a fixed AED 9,000 build divides across two weeks:
| Stage | Time | What happens | |---|---|---| | Scope and data model | 1 day | Screens agreed, entities named, integrations listed, scope frozen | | First working build | 3 to 4 days | AI writes the first draft against the frozen scope | | Human review and hardening | 2 to 3 days | Secrets, permissions, input validation, error states, edge cases | | Integrations and deployment | 1 to 2 days | CRM, WhatsApp, payment links, sheets, live environment | | Handover and fix window | 1 day | Repo in your name, credentials, short runbook, two weeks of fixes |
The invisible half of any quote lives in rows three and five. A demo that runs on a laptop is cheap. Software that survives a real Tuesday, with a bad input, a dropped API call and a user who clicks the button twice, is where the money goes. Our fixed-scope build tiers exist because that half is predictable once the scope is genuinely frozen.
Vibe coding is not the same thing as a cheap app
The honest version: AI compresses the build, then adds three cost lines that did not exist before.
Tool subscriptions are the small one. Seats for AI coding assistants run roughly USD 10 to 50 a month, about AED 37 to 185, and heavier agentic tiers go past that. On a two week build this is noise. It becomes real only when a company buys seats for eight people and nobody ships anything.
Onboarding is the medium one. Getting a developer productive inside an existing codebase costs USD 1,500 to 4,000 in lost time, and AI generated code is often harder to onboard into than hand written code because it is verbose and inconsistently structured across files. If you plan to hand the project to a different developer later, that bill lands on you.
The expensive one is security cleanup. Veracode's 2025 GenAI Code Security Report tested over 100 models on real coding tasks and found roughly 45% of the output carried vulnerabilities matching known OWASP categories; industry estimates put cleanup at 20 to 30% of project cost. Treat that number as theirs rather than ours, because we have not run that measurement ourselves. What we can confirm from our own reviews is the shape of it: the recurring failures are hardcoded API keys, missing authorisation checks on endpoints that look internal, unvalidated input reaching a database query, and error handlers that swallow failures silently. All four are cheap to catch during the build and expensive to catch after a customer finds them.
The AED 9,000 price keeps the same engineering discipline as the AED 40,000 agency quote, with the typing removed and the review kept in full.
What AED 9,000 buys, and what starts at AED 18,000
The boundary is easier than most agencies make it sound. Two questions decide the tier: does real money move through it, and does more than one kind of person log in.
| | Prototype or internal tool, from AED 9,000 | MVP, from AED 18,000 | |---|---|---| | Users | You and your team, single role | Customers, multiple roles and permissions | | Login | Simple or none | Accounts, password reset, session handling | | Payments | Payment link pasted in | Gateway integrated, webhooks, refunds, receipts | | Data | One or two sources, mostly read | Full data model, writes, history, audit trail | | Failure tolerance | You restart it yourself | Monitoring, alerts, rollback | | Timeline | About 2 weeks | Scoped per build, fixed once agreed |
Real examples of the AED 9,000 tier from work we have scoped in Dubai: a quotation calculator that ends the spreadsheet-by-WhatsApp habit, an intake form that writes straight into the CRM with the source tagged, an operations dashboard that reads bookings and payments into one screen, a clickable prototype you can put in front of ten real customers before committing to a build.
The AED 18,000 tier starts where a stranger touches it. A client portal with logins and invoices. A booking product that takes card payments and issues confirmations. A marketplace with two sides and a commission split. Anything where a bug means somebody is charged twice, and anything holding personal data belonging to people who did not agree to be your beta testers. If you are weighing which side of that line your idea falls on, how a two-week build is structured walks through the scoping day in detail.
Year one, not build day
The quote is roughly half the first year. This is the number that catches Dubai founders out, and it is the reason a cheap headline price is worth so little.
Maintenance across the industry runs 15 to 25% of build cost annually, and that is for software nobody is actively growing. Our support and monitoring sits at AED 3,000 a month, which covers uptime monitoring, dependency and security updates, and small changes. Some clients need a heavier package at around AED 6,000 a month once several integrations are live.
Hosting is smaller than people fear. A small internal tool on a modern platform runs AED 60 to 400 a month depending on database size and traffic. Model API usage is the variable one: a tool that summarises fifty documents a day costs a few hundred dirhams a month, and a customer-facing assistant handling hundreds of conversations costs more. We put these on your own accounts at cost and never mark them up, which also means you can see exactly what you are paying for.
Two Dubai-specific lines get forgotten in almost every generic cost guide. Arabic and right-to-left support adds roughly 15 to 25% to design and front-end work, because RTL is a layout rebuild rather than a translation file. UAE data protection requirements under the PDPL add roughly 15 to 20% when the product holds personal data, since consent capture, retention rules and deletion have to be built rather than promised.
| Year one, internal tool | AED | |---|---| | Build, fixed | 9,000 | | Support at 3,000 per month, 10 months | 30,000 | | Hosting and API usage | 1,500 to 4,000 | | Year one total | roughly 40,500 to 43,000 |
| Year one, MVP with users and payments | AED | |---|---| | Build, fixed | 18,000+ | | Support at 3,000 to 6,000 per month | 36,000 to 72,000 | | Hosting, API, gateway fees | 4,000 to 12,000 | | Arabic and RTL, if needed | +15 to 25% of design | | PDPL work, if holding personal data | +15 to 20% | | Year one total | roughly 60,000 to 110,000 |
Which is to say: first year lands at 1.5 to 2x the build number. A quote that skips this line is incomplete, not cheaper. Our full published price list shows the recurring side next to the build side for the same reason.
Why cheap development gets expensive
A freelancer at AED 3,000 to 8,000 will hand you something that runs. The demo works. The problem arrives about six weeks later, and it is always one of the same short list.
The repository is in their GitHub account, not yours. The database credentials are hardcoded in three files, so nothing can be rotated without breaking the app. There is one environment, meaning every change is tested in production by your customers. There is no error logging, so when it breaks you learn about it from a client. There is no migration history, so the schema cannot be changed safely. There is no documentation, so the next developer quotes you AED 12,000 just to understand what exists.
We have been called in to take over builds in this state. Most of the time the honest recommendation is to rebuild the core and keep the design, because reading undocumented code costs more hours than writing clean code from a frozen scope. That conversation is unpleasant, and it is the single most avoidable expense in this market.
The protection comes from three contract terms, not a bigger budget: the repository and every cloud account in your name from day one, a named human who reviews changes before release, and a written fix window after handover. All three are free to demand and expensive to skip.
Agency, freelancer, or fixed price
| | Dubai agency | Freelancer | Fixed-price build | |---|---|---|---| | First version | AED 40,000 and up | AED 3,000 to 8,000 | AED 9,000 tool, AED 18,000+ MVP | | Timeline | 3 to 6 months | 2 to 6 weeks, variable | About 2 weeks for a tool | | Price visibility | After a discovery call | Negotiated per person | Published | | Who owns the repo | Usually them until final payment | Often their account | You, from day one | | Code review | Yes, and you pay for it | Rarely | Human gate before release | | Common failure | Scope creep and change orders | Disappears, code rots | Scope is frozen, so new ideas wait for phase two |
The fixed-price route has a real cost, and it is worth naming: the scope freeze is strict. If halfway through week two you decide the tool also needs a loyalty programme, that is a new scope and a new price. Agencies absorb that kind of drift inside a longer engagement and bill you for it later through change orders. We refuse it upfront, which some people find rigid.
An agency genuinely beats us in three cases: a compliance-heavy build in a regulated sector with audit requirements, an integration into legacy enterprise systems where the discovery alone takes months, and a twelve month roadmap with a dedicated team. If your project is one of those, pay the AED 40,000 and do not feel cheated. If your project is a tool your operations team needs by the end of the month, three months of discovery is money set on fire.
We build our own tools this way first
Everything described here runs on our own operation before it is sold. The AI agent that sells studio rental for the SkyLight group works around the clock and writes bookings into the CRM and calendar without a person in the loop. It has closed a full deal solo, from first message to payment. The internal dashboards our team uses to watch that funnel were built the same way, in days rather than quarters.
We also keep an error log of real production incidents from that system, and every one of them feeds back into how the next build is reviewed. That log is the reason we insist on the human gate rather than trusting the model. The failures were never exotic. They were a permission check that was never written and a timeout nobody handled.
If the process you want automated already exists across tools you pay for, a custom build may be the wrong answer entirely, and connecting the systems you already run is cheaper. If you are not certain which of the two you need, the Growth Audit at AED 3,000 ranks what to build first by payback before anyone writes a line of code.
