MVP Development Company in India: The 90-Day Playbook

A good MVP development company in India will hand you a working, instrumented product in about 90 days for ₹4,50,000 to ₹28,00,000 plus GST, depending on the shape of the build. Not a prototype. Not a clickable mockup. A product a real cohort can sign up for, use unsupervised, and complain about.

Most founders who walk into a Hyderabad dev shop arrive with a 40-screen scope and a 90-day deadline. Both cannot survive. The first fortnight is spent deciding what you will not build.

Here is the playbook we run: three sprints, cost bands in INR, the features to defer, the clauses that decide who owns the code, and the five numbers investors ask for on day 90.

The short answer

MVP shape	What ships	Calendar time	DevXAI band (excl. GST)
No-code MVP	Airtable with Glide, run by hand	2–4 weeks	Under ₹1,50,000
Single-workflow web app	One user type, one core loop	8–10 weeks	₹4,50,000–₹9,00,000
Two-sided marketplace	Both sides live, manual matching	10–12 weeks	₹7,00,000–₹14,00,000
Mobile-first consumer app	One platform, Android first	10–13 weeks	₹8,00,000–₹16,00,000
Transactional MVP	UPI collections, payouts, ledger, KYC	12–14 weeks	₹14,00,000–₹28,00,000
GST	18% on IT services	—	+18%

₹9,00,000 is a little under USD 9,400 at August 2026 rates. That is what a seed-stage founder in India should budget for a first version, and it should buy a product, not a demo.

What an MVP is not

Four beliefs turn up in almost every first conversation, and each costs weeks. Naming them in week one is the cheapest hour of the project.

• Not a cheap version of the full product. A cheap version of ten features is ten broken features. An MVP is one feature that beats the manual process it replaces.
• Not a prototype. A prototype answers "can this be built". You already know it can. An MVP answers "will anyone use it twice".
• Not a design mockup. A Figma file with 30 screens is a scope document anchoring engineers to unvalidated decisions. Good MVP product design produces about eight screens.
• Not an unpaid pilot. Give the first version away to dodge the pricing conversation and you delete the only signal that matters. Charge something, even ₹99.

The working definition: the smallest thing that lets one real user finish one valuable job end to end, without you in the room, instrumented well enough to show whether they came back on day seven.

The scope-cut method: name the one workflow

Before any estimate, write one sentence in this shape: a [user] can [do the thing] and [get the outcome] without talking to us. A clinic owner publishes slots and receives a confirmed booking, unattended.

Then put every feature through three questions.

• Does removing it break the sentence? If not, it is version two.
• Can a person do it by hand for the first 200 users? If yes, a person does it.
• Does it exist only because a competitor has it? Then it is not in the 90 days.

Version two means version two, not "later in the 90 days". Founders who cannot hold that line reach day 90 with 80% of four features and 100% of none.

Asked for in week one	What ships in the 90 days	Why
Admin panel with dashboards and exports	Retool or Metabase on the same Postgres	Half a day of setup, not three weeks of internal screens
Roles: admin, manager, staff, viewer	Two roles. A boolean column, is_staff	A permission matrix costs two weeks; you have four employees
Multi-tenancy and white-labelling	Single tenant, one database	Adding tenant_id later is a two-day migration; designing for it now constrains every query
Push, email templates, SMS campaigns	One transactional email, one OTP SMS via MSG91	Nobody churned over a missing email header image
Dark mode, theming, onboarding tour	None of it	No user asks for dark mode before the feature works
Wallet, refunds, invoicing, reconciliation	One collection flow on Razorpay; refunds by hand	200 manual refunds cost less than three weeks of automation

The 90-day plan, sprint by sprint

Sprint 1, weeks 1 to 4: discovery and scope-cutting

Nothing user-facing ships in the first fortnight. Sprint 1 interviews the people who will use the thing, watches how they do the job today in WhatsApp and Excel, and rewrites the workflow sentence until both sides agree. It ends with three artefacts.

1. A data model and about eight screens, not thirty.
2. A signed list of everything deferred to version two.
3. A skeleton deployed on a real domain, on AWS Mumbai or Vercel, with auth and CI working. Empty, but real.

Sprint 2, weeks 5 to 9: build the core loop

One team, one Postgres database, one deployable. No microservices, no Kubernetes, no event bus: those solve problems you earn at scale, and adopting them in week five buys a month of infrastructure work for nothing. Working software lands on a staging URL every Friday, which is what a sane delivery process is built around.

Sprint 3, weeks 10 to 13: instrumented launch to a first cohort

Analytics events are named in week nine, so the day you have users you have data. Then a closed launch to 20 to 50 people you can name and phone, not a Product Hunt post. Weeks 12 and 13 fix what the cohort broke.

Weeks	Sprint	What exists at the end	What the founder does
1–2	Discovery	Workflow sentence, data model, signed deferred list	Produce six real users. Say no to your own ideas
3–4	Discovery	Eight screens designed, skeleton deployed, CI green	Approve designs once. Open cloud and gateway accounts
5–6	Core build	First half of the loop working end to end	Click it every Friday. Answer inside a working day
7–9	Core build	Full loop working, payments in test mode, events named	Line up the launch cohort by name
10–11	Launch	PostHog and Sentry wired, payments live, cohort onboarded	Phone everyone who signed up and did not finish
12–13	Launch	Bugs fixed, four weeks of usage data, handover done	Decide: double down, change the loop, or stop

What an MVP development company in India charges, and what moves the number

The bands DevXAI quotes assume a senior-led team of three to four people. Four things move a quote.

1. Money movement. Collecting, splitting and paying out needs a ledger, idempotent webhooks and reconciliation. That is the jump from the ₹9,00,000 band to the ₹18,00,000 band, as this guide to UPI payment gateway integration in India explains.
2. Regulated data. KYC documents, health records, anything touching Aadhaar-based verification. Consent capture, audit trails and DPDP Act, 2023 retention rules add weeks.
3. Integrations you do not control. A hospital's HIS, a bank sandbox, a GSTN filing flow. Their support turnaround sets the pace, so double the estimate.
4. Design ambition. Eight functional screens is a fortnight. A bespoke component library with motion is a month.

Add 18% GST. If you are GST-registered it is input credit rather than cost, so the timing of your registration matters. This is not legal or tax advice; confirm with your CA. The same logic for apps sits in this breakdown of mobile app development cost in India.

When you should not hire an agency at all

If you or a co-founder can write code and the product is one CRUD workflow, building it yourself in six weeks beats paying an agency ₹6,00,000. You will learn things no handover document transfers.

If your workflow is genuinely a spreadsheet with a login, use Airtable with Glide and run it manually. We have watched founders in Hyderabad spend ₹7,00,000 automating a process with eleven users. Eleven users is a WhatsApp group.

The technology recommendation that is usually wrong

Everyone will tell you to build a mobile app in Flutter or React Native. For most consumer MVPs in India, ship a fast mobile web app first. A large share of your users are on budget Android handsets with little free storage and patchy 4G, and every install is a funnel step you pay for twice. Prove demand on the web, then spend ₹8,00,000 on a native mobile product once retention is worth defending.

Equity instead of cash, and why it usually fails both sides

Founders ask most weeks, and it fails predictably. An agency's cost base is salaries, paid in rupees on the last working day of every month. Equity does not pay salaries, so an equity build gets bench capacity and loses its senior people the moment a cash project needs them. The agency gets a lottery ticket it cannot value.

A services firm on your cap table is also a line item every future investor asks about.

• What works: cash, for a scope small enough to afford. A ₹4,50,000 MVP you own outright beats a ₹20,00,000 equity build you resent.
• What sometimes works: part of the fee tied to a defined milestone, with a date and an amount in INR. That is deferred cash, not equity.
• What actually solves it: a technical co-founder. Equity suits someone still answering support tickets in year three, not a vendor whose engagement ends on day 90.

Ownership, contracts and the Indian paperwork

Source code and IP

• Assignment of all IP to your company on payment, named explicitly: source code, designs, schema, documentation. "Licence to use" is not ownership.
• The repository lives in your GitHub organisation from day one, with the agency added as collaborators. This prevents most handover disputes.
• Cloud, domain, gateway and analytics accounts in your company's name. If production sits inside a vendor's AWS account, you have a hostage, not a product.
• A written handover: runbook, environment variables, deployment steps, recorded walkthrough. Put it in the contract; nobody writes it voluntarily.

Ownership and access are the fastest filter when comparing firms; there is a longer list in this note on choosing a software development company in Hyderabad.

DPIIT Startup India recognition and Udyam registration

DPIIT recognition under Startup India is open to a private limited company, LLP or registered partnership within ten years of incorporation and under the turnover ceiling. It is the route to Section 80-IAC income tax exemption through the Inter-Ministerial Board, labour-law self-certification, the Seed Fund Scheme, and relaxed prior-turnover criteria in public procurement. Apply in month one.

Udyam registration takes minutes with PAN and Aadhaar and makes you a recognised micro or small enterprise: delayed-payment protection under the MSMED Act, collateral-free credit, tender preferences. If your MVP sells B2B in India, that changes how your invoices get treated.

What the build needs from the founder

The strongest predictor of whether a 90-day MVP lands on time is not the engineering team. It is whether the founder answers questions. One named decision-maker, replying within a working day in IST hours, not a committee.

• Three hours a week: one review call, one pass through staging, one written decision list.
• Six real users in week one and 20 to 50 in week ten. If you cannot find six people with the problem, 90 days of engineering will not create them.
• Content and data before Sprint 2: catalogue, price list, legal copy, real document templates. Placeholders hide problems until the worst week.
• A decision at day 90, against the numbers rather than your mood. Stopping is legitimate, and far cheaper than at day 400.

Founders we work with span Hyderabad, from HITEC City and Madhapur through Gachibowli and the Financial District, plus remote teams in Bengaluru, Pune, Mumbai, Chennai, Vijayawada and Visakhapatnam. Remote works fine. Unresponsive does not.

What to instrument for the investor conversation

At day 90 you will be in a room explaining what happened. Five numbers do the work, each designed in before Sprint 2 ends.

1. Activation rate. Share of signups completing the core action once, in their first session. Across the launches we have run, below 30% has meant onboarding is the problem.
2. Time to first value. Median minutes from signup to first completed core action. The number you can move in a week.
3. Week-4 retention. Of those who activated in week 10, how many acted again in week 13.
4. Core action frequency. Actions per active user per week. Flat is fine; declining is the conversation to prepare for.
5. Unit cost per completed action in rupees: gateway fee, SMS, cloud, support minutes.

The tooling is boring and cheap: PostHog for events and session replay, Sentry for errors, a Metabase dashboard for cohort tables. What not to show: downloads, signups, waitlist size, and GMV run-rates from three friendly customers. Investors in India read those as missing retention data.

Frequently asked questions

How much does an MVP development company in India charge in 2026?

We quote ₹4,50,000 to ₹9,00,000 for a single-workflow web MVP, ₹8,00,000 to ₹16,00,000 for a mobile-first consumer app, and ₹14,00,000 to ₹28,00,000 once money movement, KYC and a ledger are involved. Add 18% GST. A no-code first version can land under ₹1,50,000.

Can an MVP really be built in 90 days?

Yes, if the scope is one workflow: four weeks of discovery, five weeks of core build, four weeks of instrumented launch and fixes. Ninety days does not fit four features, three platforms, an admin panel and a permissions matrix. Any firm agreeing to that plans to be late.

Who owns the source code of an MVP built by an agency?

You should, from day one. Insist on written assignment of all IP to your company on payment, the repository hosted in your own GitHub organisation, and cloud, domain and gateway accounts opened in your company's name. List any agency libraries carved out of that.

Will a development company build my MVP for equity?

Some will, and it usually disappoints both sides. Equity does not pay monthly salaries, so equity builds get bench capacity and slip behind cash work. A services firm on your cap table also becomes a diligence question at seed. Cash works better.

Do I need DPIIT Startup India recognition before building an MVP?

Not to build, but apply early. DPIIT recognition opens the route to Section 80-IAC tax exemption via the Inter-Ministerial Board, self-certification under several labour laws, the Seed Fund Scheme and procurement relaxations. Udyam registration is free and protects B2B invoices. Check current notifications.

Start with the one workflow

Write the sentence today. A [user] can [do the thing] and [get the outcome] without talking to us. If it takes two sentences, your MVP is two MVPs and you have to pick one.

Then list every feature in your head and put each on the deferred side unless removing it breaks that sentence. Most founders we sit with find well over half their scope is version two.

Send the sentence and the cut list to DevXAI and you will get back what it costs, how long it takes, and whether hiring an agency is right. Or start from the work: recent projects and the web platform engineering practice.