MVP development services, sized to the question you need answered
From a $2.5K validation build in two weeks to a $200K multi-product MVP in five months. The bands below come from our own estimates and delivered projects, not from a pricing page.
What an MVP costs, and how long it takes
Validation build: $2.5K–18K, 2–6 weeks
50–300 hours on top of a prototype that already exists, or a narrow proof of concept on real data. One question answered with working software — will restaurants post campaigns, will a parser read these purchase orders — before anyone pays for the rest.
Lean web MVP: $10K–20K, 4–8 weeks
170–375 hours for one core journey, an admin and notifications. A marketplace for community-solar projects launched at this size: listings, bids, offers and commitments for 283–352 hours including design, PM and QA, and has kept shipping releases since.
Standard web or mobile MVP: $35K–105K, 8–16 weeks
550–1,150 hours. The usual shape for a product with two sides, a mobile app, or a prototype that has to become secure before real users touch it: roles, row-level security, file handling, QA and release operations.
Multi-product or integration-heavy MVP: $109K–200K, 16 weeks to 5 months
1,300–3,300 hours. Two products on one plan, or one product whose critical path is someone else's API — a booking system, a CRM, a payment provider — where most of the risk sits in the integration rather than the screens.
MVPs we have built
Prototype to production for a collegiate sports platform
A platform for name, image and likeness deals, built on top of the client's Lovable prototype rather than thrown away and rewritten. Planned as five two-week sprints including Sprint 0; staging two weeks after the plan was signed, production three months later. One lead developer, a second from week two. FastAPI, React and PostgreSQL, with an AI coach on GPT-4o-mini streaming answers inside the live product.
A HIPAA-aligned member app in about eleven weeks
A health assessment for a membership wellness center: eight domains, a composite score built from 89 atomic tests, 35 functional requirements. Half a PM and one full-stack engineer, with design, QA and DevOps as needed, took it from a signed statement of work to launch, on Next.js, NestJS, PostgreSQL on RDS and ECS Fargate, inside an AWS configuration covered by a signed BAA. AI-generated plans were left out of the first statement of work on purpose and shipped later as their own phase.
A voice agent MVP in about three months
An AI agent that answers inbound calls for home-services businesses, collects the lead's details and hands over to a person. Kickoff in mid-March 2026, first demo in late April, inbound calls in production on June 9 and the MVP delivered on June 26: 245 tickets, 751 unit tests in the voice service, against an estimate of 420–670 hours. More on how it works on our agentic AI development page.
A B2B marketplace that kept going after V1
Solar project developers meet acquisition companies: capacity listings, bids, offers, negotiations, commitments and an approval model. Scoped at $12K–14K for the early version and about $18.4K for the full 307-hour PRD, with a first delivery window of four to six weeks. It is live, with release 1.11.0 shipped in July 2026, and we led the product design as well as the build.
How an MVP runs, from brief to hypercare
- Days 1–3: estimate
Day one, estimators are assigned to the brief. Day two, clarifying questions, or a first version with its assumptions written down. Day three, a rough estimate and a call to go through it. Every estimate is stored with its version history, so a later number can be traced back to what changed.
- Weeks 1–3: discovery, fixed price
Product and UX discovery with a clickable prototype in about three weeks, or a technical validation of the risky integrations in one to two. Critical dependencies are tested before anything is promised: on a booking-integrated mobile app, the booking system's limit of 60 API calls a minute is one of the things validated before the scope is locked.
- Sprint 0: one to two weeks
The technical foundation and a scope lock, or a proof of concept on real client content. The PRD fixes the V1 goal, what is in and out, assumptions, acceptance criteria and the open questions with their effect on the estimate.
- Build: 4–16 weeks
Two-week sprints with a demo at the end of each, a weekly sync or written summary, and a project workspace the client can see. Each release is something a user can use, not an internal checkpoint.
- UAT and release: one to two weeks
A QA week, full regression of at least four hours, a readiness review and a go/no-go from the client. For mobile, store submission is planned two weeks before launch.
- Hypercare: two to four weeks
Priced separately rather than buried in the build, so it is visible. For a sixteen-week mobile MVP it is planned at about 160 hours.
What keeps an MVP on scope
A PRD with an explicit V1 goal
Must-haves, exclusions, assumptions and acceptance criteria, with priorities marked MUST / SHOULD / NICE or tiered A/B/C. When a budget is tight, the tiers decide what moves to V2 instead of an argument mid-sprint.
Estimates sized in work packages
Features are sized as XS 40, S 80, M 120, L 160 or XL 200 hours, with base and high columns kept separate from the calendar plan. That is what lets one product be priced at four sizes, and lets a client see exactly which line a change touched.
Change control, because we have been on the wrong side of it
One earlier MVP, estimated at $55K–85K in 2023, finished above its estimate after a run of scope changes. Since then every change is recorded with its hour impact, and we track refinement completion above 90%, release predictability above 85%, bug rate below 5% and scope deviation below 10% on every project.
Sized to the hypothesis, not to a template
In October 2026 we priced one restaurant and influencer marketplace four ways on top of the prototype it already had: a $5K validation build in 2–3 weeks proving campaign to application to selection, $15K for the lean version, $18K with payments, and $52K for the full MVP. The founder decides how much certainty to buy before paying for the rest.
AI in the delivery, priced in
Engineers work in Cursor, Claude Code, Codex and Copilot, with human review on every change. Our estimates show the effect rather than promise it: the same real-estate catalog scope came out at 617 hours with AI-assisted development against 714 without. v0, Lovable and Figma Make are for prototypes only; production still needs role-based access, payments and proper handling of personal data.
Who builds it, and on whose clock
Where the team sits
More than half our engineers now sit in Latin America, with most of the rest in Eastern Europe. On an MVP the overlap is the schedule: a demo every two weeks only works if the founder can watch it and answer the questions it raises the same day.
How people get here
Hiring runs through the same funnel as every other stack here — roughly one hire per 130 applications on the front end and 157 on the back end. There is no cheaper tier of engineer for MVP work, because an MVP that works is the one that becomes the product.
Frequently Asked Questions
How much does an MVP cost?
How long does it take to build an MVP?
What team builds it?
We already have a prototype from Lovable or v0. Do we start over?
Fixed price or time and materials?
Can the MVP include AI?
What happens after launch?
Ready to get started?
Contact us at hello@scrumlaunch.com or fill out the form
