Pricing
What things cost, and what makes them cost more.
Ten of the twelve studios we surveyed while building this site publish no prices at all. Here are ours, along with the things that genuinely move a quote up or down.
Shark Labs publishes fixed price ranges for every service. A marketing site costs $1,500–6,000, a web application with accounts and data $6,000–18,000, a mobile or web MVP $6,000–20,000, and an AI feature added to an existing product $3,000–8,000. Monthly retainers are $900, $2,500 and from $6,000. Every project starts with paid discovery of $400–1,800, which is credited against the project price if you go ahead.
Project pricing
Ranges, not “contact us for pricing”.
Web and apps
| What | Price | In ₹ | Timeline |
|---|---|---|---|
| Marketing site | $1,500 – $6,000 | ₹1.25L – 5L | 3–6 weeks |
| Web application with accounts and data | $6,000 – $18,000 | ₹5L – 15L | 6–14 weeks |
| Mobile or web MVP, including backend | $6,000 – $20,000 | ₹5L – 17L | 8–16 weeks |
| Multi-role product with payments and admin | $20,000 – $35,000 | ₹17L – 30L | 14–24 weeks |
AI
| What | Price | In ₹ | Timeline |
|---|---|---|---|
| AI feature into an existing product | $3,000 – $8,000 | ₹2.5L – 7L | 3–6 weeks |
| RAG system over one corpus | $4,000 – $10,000 | ₹3.5L – 8.5L | 4–8 weeks |
| Agent with tool access | $8,000 – $20,000 | ₹7L – 17L | 6–12 weeks |
| AI-first product or MVP | $8,000 – $25,000 | ₹7L – 21L | 8–16 weeks |
Before and after
| What | Price | In ₹ | Timeline |
|---|---|---|---|
| Discovery (credited against the project) | $400 – $1,800 | ₹35k – 1.5L | 2 days – 1 week |
| Codebase audit / takeover assessment | $900 – $1,800 | ₹75k – 1.5L | 1 week |
Retainers
After launch, priced the same way.
Of the twelve studios we surveyed, one had a dedicated retainer page and none published a price. Full retainer details.
Keep the lights on
$900 / month₹75,000 / month
- Hours
- ~10 hours
- Response
- 2 working days
- New features
- Not included
Monitoring, security patching, dependency updates and bug fixes. Nothing degrades. No new features.
Keep it moving
$2,500 / month₹2.1 lakh / month
- Hours
- ~30 hours
- Response
- Next working day
- New features
- Included
Everything above, plus feature work, AI model migrations and eval upkeep. The tier most live products want.
Dedicated
From $6,000 / monthFrom ₹5 lakh / month
- Hours
- ~80 hours
- Response
- Same day
- New features
- Included
Effectively an embedded part-time engineer, with weekly calls and priority incident response.
Unused hours roll over for one month. Thirty days' notice, no minimum term.
How we quote
Every project is fixed-price, quoted after a short paid discovery rather than from a conversation. That order matters: a number produced before anyone has looked at the code or the data is a guess, and guesses get corrected in your direction later.
- 01A free scoping call, 45 minutes. Sometimes this ends with us saying the thing you're describing doesn't need us, which is a real outcome and not a sales technique.
- 02Paid discovery, $400–1,800 depending on the project. We go into the actual code and data and produce a written plan: what gets built, what it costs, when it lands.
- 03A fixed price and a date. Not an hourly estimate that drifts. If we get the estimate wrong, that's ours to absorb — which is why discovery exists.
- 04Discovery is credited against the project price if you go ahead. If you don't, you keep the plan and can take it to another studio.
What actually moves the number
Two projects with identical descriptions can differ threefold. Here's what does it, in order of impact.
| Factor | Effect |
|---|---|
| The state of your data | The single biggest driver on AI work. Clean and accessible sits at the bottom of the range. Spread across four systems with no shared identifier can double it before any model work begins. |
| Permission complexity | "Everyone sees everything" and "each user sees only what they're entitled to" are weeks apart in a retrieval system. |
| Whether it takes actions | A system that answers is well-understood. A system that acts on your data needs permission design, audit logging and confirmation flows. Roughly doubles the surface. |
| Integration count | Each external system is its own auth, rate limits, failure modes and sandbox. Three integrations is not three times one — it's more. |
| Where the model runs | A hosted API is the cheap path. Self-hosting an open-weight model because data can't leave adds infrastructure and ongoing operations. |
| Design | Working from your existing design system is fastest. Designing from scratch adds one to three weeks. |
| Deadline pressure | A date that requires reordering other work carries a premium. We'd rather tell you that than quietly miss it. |
What we don't charge for
- The scoping call. Forty-five minutes, no obligation, and you'll get a straight answer about whether this is worth doing.
- Estimate overruns. Fixed price means fixed. If we misjudged the work, that's ours.
- Licences and margin on your infrastructure. Your cloud, model provider and third-party accounts are yours, billed to you directly. We don't resell hosting with a markup.
- Handover. Documentation and a walkthrough with your engineers are part of every project, not an upsell.
How payment works
- Discovery is paid up front. It's small and it's the first thing we do.
- Projects are 40% to start, 30% at the agreed midpoint, 30% on delivery. Longer projects split into more milestones so neither side is ever far ahead of the other.
- Retainers are billed monthly in advance, with thirty days' notice to cancel and no minimum term.
- Currency. USD, EUR, GBP or INR. Indian clients are invoiced in INR with GST as applicable; overseas clients are invoiced in their own currency with no GST.
Where we sit against the market
Honestly, so you can calibrate rather than take our word for it.
| Typical range for one AI feature | |
|---|---|
| Large consultancy | $50,000 – $250,000+ |
| Established mid-size agency | $15,000 – $60,000 |
| Shark Labs | $3,000 – $8,000 |
| Freelance marketplace | $1,000 – $5,000 |
We're cheaper than an agency because there is no sales team, no account layer and no office. We're more expensive than a marketplace because the eval suite, the guardrails, the cost model and the documentation are in the price rather than absent from it — and because we're still here in six months.
If the marketplace price is what your budget allows, take it. It's a legitimate option for a well-specified, low-stakes piece of work, and we'd rather say so than pretend otherwise.
The number most quotes leave out
Running cost. An AI feature has a monthly bill that scales with usage, and over three years it typically dwarfs the build. For a feature at around five thousand active users, the build is often under 5% of the three-year total.
We model this during discovery — calls per user, tokens per call, cache hit rate, and what it becomes at ten times your current volume. What an AI feature costs, with the arithmetic shown walks through it in full, including the three levers that cut it most.
FAQ
Questions about pricing
Why do you publish prices when most agencies don't?
Because an unpublished price is quoted at whatever the buyer looks able to pay, and because the alternative wastes both sides' time — buyers whose budget doesn't reach us find out on the third call instead of the first minute. Publishing costs us the ability to price-discriminate. It buys better-qualified conversations, and we think that's the better trade.
Are these prices fixed or estimates?
The ranges on this page are indicative. The number you get after discovery is fixed — a specific figure, in a written scope, that doesn't move unless you change what's being built. If we misjudge the effort, that's ours to absorb, which is precisely why we do paid discovery before quoting.
What if we only have part of the budget?
Say so on the first call. Most projects have a smaller true first version, and finding it is a better conversation than a proposal you can't accept. What we won't do is quote the full scope and then quietly deliver less of it, which is the usual way this gets resolved.
Do you charge differently for Indian and overseas clients?
No. Same scope, same price, converted at the prevailing rate. Indian clients are invoiced in INR with GST as applicable, overseas clients in their own currency without it. There's no offshore discount and no Western premium — we found both practices while researching this market and neither survives being written down.
What's included in the price that I should check others include?
An evaluation suite for anything AI, monitoring and alerting, a written cost model, handover documentation, and a walkthrough with your engineers. Those are the line items most often absent from a cheaper quote, and each one is something you'll pay for later rather than avoid.
Do you do equity or revenue share instead of fees?
No. It's a reasonable thing to ask and the answer is that a studio taking equity across several clients is running a fund badly. We'd rather quote a number you can decide on.
Next step
Get a real number.
Forty-five minutes on a call and we can usually give you a range on the spot. A fixed price takes a few days of paid discovery, and you keep the plan whichever way you go.