Every price, one place.
Landing Page
Be found and understood by the people who matter.
Starting at $2,500
Includes one year of hosting and 30 days of post-launch monitoring.
What the floor includes
- One page, six to eight sections, built to be understood in the first ten seconds.
- Responsive from phone to desktop, with one motion pass that guides the eye instead of decorating the page.
- A working way to reach you: a contact email, or a scheduling link through a service like Cal.com or Calendly.
- The search foundation search engines and AI assistants need to read your business correctly.
- Your domain connected, deployed live, and hosted securely.
- Two rounds of revisions before launch.
- One year of hosting.
- 30 days of post-launch monitoring.
Not a dead end
The build is real code, not a template or a page builder. It sits on the same foundation as our larger work, so more pages, a blog, or a real content system later are extensions of what you already own rather than starting over.
What counts as a page, what makes one cost more than another, and how anything you add connects back to the main one. Everything beyond the base page lives in the same navigation, as a real destination a visitor can click to.
- The page itselfIncluded
- One page, six to eight sections, one continuous scroll. There is a navigation bar, and its links are anchors that move you down the same page rather than off to separate destinations. A section is a self-contained band of that page: the hero, what this is, the offering, the proof, the human behind it, how to get in touch. Which ones you get depends on what your business actually needs, ordered the way a stranger needs them rather than the way you'd explain it to someone who already knows.What this means
- Simple page$150–350
- A page that is mostly words. About, FAQ, privacy, terms. One layout, read top to bottom, nothing in it that reacts to you.What this means
- Standard page$300–600
- A page with repeating structure. A services list, a menu, a portfolio grid, a team page. Every item needs its own treatment, and all of them have to hold together as a set once they're side by side.What this means
- Interactive page$800–1,800
- A page that does something rather than shows something. A booking system embedded in it, a form that runs across several steps, filtering across a set of items, a live map. There's logic behind it and not just layout, and logic has to be built and then tested against the ways people actually use it.What this means
- How many you can add
- Four beyond the base page, maximum. Past that it stops being a landing page and turns into a website, which is a different project with a different shape and a different price.That ceiling isn't a policy we bend, because the promise of this tier only holds at this size. It's also not a dead end. The build sits on the same foundation as our larger work, so growing later is an extension of what you already own rather than starting over.What this means
What you bring, what we make, and what happens when you have neither.
- What you supplyIncluded
- Copy, or the points you want the copy to make. Your logo and whatever brand assets exist. Photography. The practical facts of hours, location, and how to reach you. Access to wherever your domain is registered. When all of that is sitting ready on day one, the timeline holds. When it isn't, that is almost always the thing that moves it.What this means
- Copy written for you$800–1,500
- If you'd rather not write it, we interview you, draft it, and revise it with you until it sounds like you rather than like a website. Add about a week to the schedule.What this means
- Direction, if you don't have oneIncluded
- Plenty of businesses arrive with no settled type, color, or feel. We work one out with you and put it in front of you as a style tile before a single section gets built. You approve that direction while it's still a small thing to change, which is how the look gets settled early instead of argued about the week before launch.What this means
- Imagery1–2 images$1003–5 images$200–500
- If you don't have photography, we can generate imagery for the page, or tell you what kind of shots would work if you'd rather take your own. Generated imagery gets treated as real production work: art-directed, iterated, and matched to the type and color around it. It is not prompted once and dropped in.Both routes take real studio time, which is why the working allowance below matters.What this means
- The working allowance$100–150/hr
- Discovery, direction, imagery, and copy all draw from one fixed block of studio time, set in your brief before anything starts. It's a specific number and you'll know what it is. If something you ask for would run past it, you hear that before the work happens rather than in an invoice afterward.What this means
Where ordinary motion ends and a video build begins. That line is what changes the shape of the project.
- The motion passIncluded
- Included in every build. How things enter, how the page answers a scroll, how one section hands off to the next. Motion that carries meaning and walks the eye down the page instead of decorating it. It gets designed once, for the page as a whole, rather than bolted onto each section as an afterthought.What this means
- Scroll-driven video$800–1,800
- An entirely different thing: video or a frame sequence that scrubs forward and back under your thumb as you scroll, the way a product reveal does. It needs frame extraction, re-encoding, separate variants for phone and desktop, and a still fallback below a certain screen size. That's its own build, not a bigger motion pass.If you've seen it on a car or phone launch page and want that, this is the thing you're describing.What this means
- Why video changes things
- Video is heavy, and what comes out of a camera or a generator is never what should be sent to a browser. It's routinely five to ten times larger than it needs to be. Compacting it is real work: cutting frame rate, re-encoding, serving something smaller to a phone on a train than to a laptop on office wifi. Even done well, a page carrying real video moves far more data on every single visit, and that is what lands it in a different hosting tier.What this means
Where the page lives, who owns what, and how you leave if you ever want to.
- What hosting coversIncluded
- We host on our own infrastructure rather than handing you an account and a password to look after. That covers deployment, uptime, and keeping the page live and loading securely. Security is handled the moment your domain is connected properly, so it isn't a task sitting on anyone's list. The first year comes with the build.What this means
- Standard hostingIncluded
- For a normal page: text, photography, the motion pass. The overwhelming majority of builds sit here and never move off it.What this means
- Media-heavy hosting$250–450 one-time, for your included yearCovers real image and video optimization work, plus room for up to around 50,000 visits a month with no surprise costs.
- For pages doing real work with images, or carrying one well-optimized scroll sequence. A gallery, a reel, a portfolio with actual weight to it. A single sequence, properly compressed, still serves fine from standard infrastructure. It simply moves more data month after month, which is why it sits one tier up from Standard hosting.What this means
- Cinematic hosting$600–1,200 one-time, for your included yearCovers the separate storage architecture your video needs, plus room for up to around 15,000 visits a month. Past that, you'll hear from us before it becomes a bill, not after.
- For pages built around several scroll sequences, or footage too dense to compress down without wrecking it. Past a certain weight the page stops being servable from standard infrastructure at all. The heavy assets move to storage built for exactly that job while the page itself stays where it is. Different architecture, not a bigger allowance.What this means
- Connecting your domainIncluded
- Wherever your domain is registered, whether that's Squarespace, GoDaddy, or somewhere you set up years ago and haven't logged into since, we'll need access to its settings so we can point it at our hosting and get your site live there. You keep ownership of the domain and the registrar account the entire time. We never register it in our name and we don't hold it.Buying and renewing the domain stays yours. It's a few dollars a year and it should stay in your name.What this means
- LeavingIncluded
- If you want to move, you take the code and the domain records with you, on request, at no charge. There's no lock-in and nothing to unwind.What this means
How people reach you, and the point where a form stops being simple.
- The contact pathIncluded
- Included: one working way to reach you. A contact email, or a scheduling link through something like Cal.com or Calendly if you'd rather people just book the time themselves.What this means
- When a form gets biggerPriced as an Interactive page
- Several steps. Questions that change depending on earlier answers. Routing to different people based on what someone picked. Anything that has to write into a system you already run. All of that is an interactive page rather than a contact path, and it gets scoped as one.What this means
Two genuinely different levels of control, depending on how much you expect to change.
- Self-serve editingIncluded
- Log in, change the text, swap an image on a section that already exists. This is what keeps your hours, your prices, and your copy current without anyone having to call us. What it won't do is add new kinds of pages or rearrange the structure. The page keeps the shape it was designed as.Set up at handoff rather than during the build, for technical reasons on our end.What this means
- Content EngineSee Content Engine below
- For when the site is genuinely going to keep growing. New items added over time, several people editing, a proper media library, things scheduled to publish while you're asleep. That's a real system rather than an edit button, and it gets scoped on a call, because what it needs to hold varies enormously from one business to the next.What this means
What we build in, and the point where a specialist takes over.
- The search foundation
- Included in every build, never a line item. Page titles and descriptions. The preview image that appears when someone drops your link into a group chat. Proper heading structure, alt text, a sitemap, structured data that search engines and AI assistants can both read, fast loading, and analytics wired up so you can see what's actually happening. This is the work that makes you findable and describes you correctly once someone goes looking.What this means
- Search strategy
- Not included, and deliberately so. Keyword research, local search and your Google Business listing, getting cited by an AI assistant when someone asks it a question in your category, earning links from other sites, publishing on a schedule. Almost none of that happens on your page. It's about what the rest of the internet says about you, which is a different job with its own practitioners. We bring in a specialist for it as a separate conversation when you're ready for one.The foundation makes you findable. Strategy makes you found first. They're different jobs.What this means
What happens when you need something changed six months from now.
- Care blocks$400–600 per block
- Blocks of studio hours you buy when you actually need them. No subscription, no card sitting on file, nothing quietly charging you in a month where nothing needed doing. Work starts within three business days of you asking, and hours you don't use stay good for a year.What this means
- If it becomes regularScoped with you
- If you find you're changing things often enough that buying blocks one at a time stops making sense, that's an ongoing arrangement and we'll scope it together.What this means
Beyond this tier
- Content Engine
- For people who will keep publishing: a blog, ongoing posts, a real content system rather than an edit button on marketing copy. Adding pages is a different thing, since a page is set once and revised twice. This is content that keeps arriving, and it gets scoped as its own conversation.
- Light commerce
- Selling a handful of products, or taking orders directly rather than pointing people at someone else's service. Scoped as its own conversation.
Prototype
Prove the core interaction works before you build it.
Starting at $4,000
Includes the repository, a one-command local setup, and the mechanic verified live with you.
What the floor includes
- The core mechanic, built for real and running, not simulated or faked for a demo.
- A written diagnosis naming the single riskiest unproven assumption in your idea, delivered before a line of code gets written.
- A repository you own, with one command that brings up the app and its database on a machine you control.
- Rounds of tuning on how the mechanic actually behaves: timing, edge cases, feel.
- An honest account of what carries forward into a real product and what was always going to be thrown away.
- Any third-party usage costs estimated and capped up front, billed to you at cost.
- The mechanic verified live, together, with you watching it run in a browser.
At this price it is deliberately plain: black and white, default styling, no design pass. Everything the mechanic needs to work is there. Everything it needs to look finished is not, and that's a choice you get to make separately rather than one we make for you.
Where it lives, once it's proven
Every engagement gets proven, watched live, together. That's never a separate line item, and it's never something you pay extra for, whether the mechanic ran entirely on your laptop or needed a real server just to exist at all. Whatever it took to reach that point is already folded into the level price above. Proven isn't a choice you make. It's the floor every single build already stands on, no matter which level it's in.
What you're actually choosing, past that floor, is whether anyone besides your own team ever gets to reach it, and if so, who.
| State | What it means | Price |
|---|---|---|
| Shareable | Deployed at a real link. No accounts, no sign-in, no data collected. Someone just clicks and tries it. Stays live 90 days. | +$1,000–2,000 |
| Unsupervised strangers | Everything Shareable already is, plus: real strangers, alone, nobody from either team watching. Needs light sign-in, usage limits, sessions that expire cleanly. | +$2,000–4,000 |
How it looks
Whether this matters is entirely your call, not a rule about who's going to see it. Plain works fine even in front of someone else, if that's genuinely how you want to show it. If you'd rather it feel a little more considered before anyone sees it, whatever your reason, that's what Presentable buys.
| State | What it means | Price |
|---|---|---|
| Plain by default | Black and white, default spacing, nothing decorative. The mechanic is the point. | Included |
| Presentable | A pre-packaged theme, built on Tailwind and styled shadcn components, applied as-is. No design round, no revision loop. | +$800–1,600 |
Custom design and visual revision are their own conversation, not something this tier covers.
How complex the mechanic is
Two different kinds of things push this number up, and every example from here on names exactly which ones apply to it. Nothing below is priced by vibe. It's priced by which of these are actually true for that specific idea.
Infrastructure the mechanic can't run without:
- Needs an always-on database: Something has to remember state between visits, not just show a snapshot.
- Needs real user identity: The system has to know which of two or more people is doing what, not just sit behind one shared password.
- Needs a live, always-open connection: The screen has to update the instant something happens somewhere else, not just when it's refreshed.
- Needs a server standing between your app and an outside service: Something has to hold a key or a credential safely and make the call on your app's behalf.
- Needs to run on its own schedule, whether or not you're there: Something has to wake up and do its job automatically, at six in the morning or once an hour, independent of you ever opening the app.
How complex the logic itself is, independent of any of that:
- The logic itself is genuinely hard to get right: Real math, real constraints, real edge cases, even if nothing outside the mechanic ever needs to know about it.
- The AI is making a real judgment call: Deciding something, not just displaying, sorting, or transcribing.
- More than one kind of input has to work together: Voice, image, video, or text, combined, not handled one at a time.
- Something has to generate in real time to match something live: Video syncing to speech as it's being spoken, for instance, not something pre-made.
- Generated output has to match something real: A real photo, a real space, a real object, not just something plausible-looking made from scratch.
Level One, alone on your machine$4,000–6,000
The real line isn't whether a server or a database exists anywhere. It's whether we have to stand one up and host it ourselves. If whatever you're connecting to already runs its own backend, its own database, even its own AI, that's fine, that's theirs to run, not ours. What actually pushes something out of this level is whether anything has to be reachable from outside your own machine: a public link, a callback some outside service needs to reach back into, a process that has to stay running and answering requests whether anyone's watching or not. A one-shot call out to someone else's service and back, even a paid one, doesn't need any of that.
Reads a folder of your own CSV exports, sales reports you download by hand, and turns them into charts and trends, refreshing whenever a new file gets dropped in. Nothing outside your own machine, nothing to protect.
What’s driving this price: nothing beyond the floor.
Drop in a recording, a meeting or a lecture, and it calls out to a transcription service and a language model to pull out the key points and action items, then hands you the result. Nothing here has to be hosted anywhere, one call out and one response back, run straight from your own machine. Getting the AI to actually return that consistently, organized the same way every time instead of a different shape each run, takes real work writing and testing the instructions it's given. The tokens and transcription minutes it burns through in real use are a separate, ongoing cost, estimated up front and billed to you at cost, on your own account, not through anything of ours.
What’s driving this price: The logic itself is genuinely hard to get right Usage is billed separately, on your own account.
You feed it historical stock prices and a rule, buy when the price drops five percent in a day, sell when it recovers three percent, and it simulates that rule across years of data. Still nothing outside your machine. What pushes the price up is getting the simulation logic right: walking through years of data in order, handling a rule that triggers twice in one day, getting the math right the first time.
What’s driving this price: The logic itself is genuinely hard to get right
One thing that looks like Level One and isn't: anything touching a real bank account. Read the transactions, sort them, show where the money went, sounds simple. In practice it needs something reachable from outside your machine to handle the login handoff safely, which jumps it out of this level entirely, regardless of how simple the logic looks.
Level Two, the state can't live on just one machine, one sitting$6,000–8,000
Three different things push a mechanic out of Level One and into needing a real server behind it. One is the pattern already covered: more than one person has to see the exact same live state change at the same moment. A second, just as common, doesn't involve anyone else at all. One person's own data has to survive past a single sitting and follow them from device to device, their phone in the morning, their laptop that night, and it has to stay provably theirs and nobody else's. A third: the mechanic has to run on its own, on a schedule, whether or not anyone's machine is even on. None of the three is about how hard the logic is. All three mean the same underlying thing: the state, or the process itself, can't live only on whichever machine you happen to be using right now.
An always-on database and real identity behind it are the actual floor of this level, present in every example here. A live, instant connection is an additional piece on top of that floor, needed only when someone, or your own devices, has to see a change the moment it happens rather than the next time something loads.
Log which exercises you did, how many reps, how much it hurt, right after each session, usually from your phone. Weeks later you pull it up on your laptop to see whether the trend is actually improving before your next appointment. Nothing here needs to update in real time. What it needs is to still be there, accurately, whichever device you open it from, days or weeks after you logged it.
What’s driving this price: Needs an always-on databaseNeeds real user identity
Two or three people working the sales floor at once. The moment one sells the last unit of something, everyone else's screen has to reflect it immediately, or the next customer gets promised something that's already gone. The logic is close to nothing: add one, subtract one, show who changed what.
What’s driving this price: Needs an always-on databaseNeeds real user identityNeeds a live, always-open connection
Every morning before the shop opens, it checks a list of competitors' product pages, pulls their current prices, and lays them out in one place so the owner can see who's undercutting them before customers walk in. It runs on its own at six in the morning whether the owner's laptop is open or not, with results waiting whenever they check, phone or laptop. If a competitor's page fails to load, or changes enough to break the read, that gets flagged for the owner rather than guessed at, and the rest of that morning's report still comes through.
What’s driving this price: Needs an always-on databaseNeeds real user identityNeeds to run on its own schedule, whether or not you're thereThe logic itself is genuinely hard to get right since a broken source has to get caught and flagged, not mistaken for real data.
That's organizing what got scraped. The moment it has to actually judge what it found, deciding which change is worth flagging, ignoring the noise, that crosses into Level Three instead, the same line that separates the voice memo transcriber from the live voice agent.
A host's tablet and servers' phones show the same live picture of every table: seated, ordered, needs bussing, ready again. A table can't skip straight from seated to ready again without passing through the states in between, enforced correctly across a dozen tables at once.
What’s driving this price: Needs an always-on databaseNeeds real user identityNeeds a live, always-open connectionThe logic itself is genuinely hard to get right
Level Three, real judgment, real integration, on the same foundation$8,000–12,000
Everything here still needs the server and real identity from the level below. What separates the examples is which of the harder difficulty tags stack on top, and how many.
Reads every incoming email, decides how urgent it actually is, drafts a reply for the owner to review and send. Never sends anything on its own.
What’s driving this price: Needs an always-on databaseNeeds real user identityThe AI is making a real judgment call
A landscaper takes a photo of an actual yard, feeds in reference photos of gardens the customer likes, and gets back a realistic reimagining of that specific yard in that style, alongside a real cost estimate.
What’s driving this price: Needs a server standing between your app and an outside serviceMore than one kind of input has to work togetherGenerated output has to match something real
A customer talks to it out loud, it has to listen while still deciding what to say and handle being interrupted, pulling answers from the company's own documentation, with a realistic avatar generated in sync to speak the answer back.
What’s driving this price: Needs an always-on databaseNeeds real user identityNeeds a live, always-open connectionThe AI is making a real judgment callMore than one kind of input has to work togetherSomething has to generate in real time to match something live
Past that, scoped with you
The complexity here can't be priced from the outside, because it depends on specifics nobody can answer without a real conversation first.
Dozens or hundreds of people scoring points and updating a shared leaderboard at the same moment. Every tag from Level Two already applies. What can't be priced in advance is how well it holds up under real concurrent load, which depends entirely on how many people, how fast, and how much lag the game can tolerate.
Several AI agents each own a different job and have to hand work to each other, retry cleanly on failure, and not get stuck when two of them disagree. How many agents and how they resolve a disagreement changes the shape of the whole build.
Two tags you won't see priced anywhere on this page, and why. Needs a vector database and AI executes a real financial transaction on its own both get redirected instead of quoted, no matter how they're framed. The first is a specialized discipline of its own, evaluating and tuning retrieval quality, closer to a research loop than a build task. The second means real money moving with nobody confirming first. Either one comes up, and we say so plainly in the first free conversation, before discovery ever starts.
If you could search "how to integrate X" and be done, that's plumbing, not a prototype. It's a prototype when nobody can tell you in advance whether the interaction will actually work, or feel right.
How the choices stack
- Level One floor, still just proven, plain: $4,000. Nothing added, because proven never costs extra.
- Level One floor, made shareable: $4,000 + $1,000–2,000 = $5,000–6,000
- Level Two floor (exercise tracker), still just proven, plain: $6,000. Already above the Level One ceiling, because the database and real identity behind it were never optional here, not because anything got added on top.
- Level Two ceiling (table board), made shareable and presentable: $7,750 + $1,000–2,000 + $800–1,600 = $9,550–11,350
- Level Three top (voice and avatar), unsupervised strangers, presentable: $11,500 + $2,000–4,000 + $800–1,600 = $14,300–17,100
When the 90 days are up
Shareable hosting runs 90 days from delivery, and we'll reach out well before it lapses so nothing goes dark in the middle of a conversation. From there you take the repository with you and run it yourself, or it's time for the MVP conversation. We're not in the business of maintaining this for you, and if you want someone who is, we'll point you at them rather than quietly become them.
Discovery
Real diagnostic work, paid, not a sales call wearing a different hat. Ends in a short written document: the problem stated plainly, what's in and out of scope, a sketch of how it'd be built, the biggest risks, priced options with a plain-English recommendation.
| Level | Standalone | Credited if we keep going |
|---|---|---|
| Level One | $1,000 | $500 |
| Level Two | $1,500 | $750 |
| Level Three | $2,000 | $1,000 |
| Scoped with you | $2,500 and up | not yet determined |
Beyond this tier
- Moving up to an MVP
- Real strangers, real accounts, data collected off the back of it, a system built to run unsupervised. That's a different project with a different shape, not a bigger version of this one.
- Design authored for you
- Applying a ready-made theme is part of this tier. Designing something original — your own type, your own layout, rounds of visual revision — is real design work with a real designer's time behind it, and it gets scoped separately.
- Work we'd send you elsewhere for
- Building the engine underneath rather than composing what exists. Training a model from scratch. Native mobile using device APIs. Advanced 3D or motion production. Systems needing very high concurrent load on day one. And anywhere we'd have to own and certify the compliance ourselves rather than compose tools that already carry it.
MVP
Ship the smallest version people can actually use.
Starting at $13,000
Includes the application built and deployed, a real handoff with documentation and a ninety-day warranty, and the infrastructure it runs on, named upfront.
MVP means Minimum Viable Product, not the finished app: the smallest real version that can actually answer whether people want this. Most of what we build here starts as a Prototype with us first: proving the core mechanic before committing to the full build, the same way the examples below show.
What the floor includes
- The application itself, built and deployed live, scoped to what your first real users actually need rather than a wishlist of what they might need someday.
- A layered architecture, view, business logic, and data kept separate, so a screen, a flow, or a whole feature can change without tearing up the foundation underneath it.
- A real handoff: documentation kept current as it's built, your full project history in a format you can actually keep, and a ninety-day warranty against anything that doesn't match the spec.
- The infrastructure your app actually runs on: hosting, database, authentication, email, error tracking, and analytics, named and set up rather than left for you to assemble.
- Privacy and security basics built in from day one: cookie consent, encryption, a real authentication provider, and a real way for a user to delete their own data.
- A named, bounded advisory option after launch, for the times you want us back in the room without signing up for anything ongoing.
The Levels
The dividing line here isn't dollar value or subject matter. It's whether a mistake in the system surfaces immediately and gets caught, or hides and compounds silently until it's already done damage. A failed reminder email is loud; a race condition in a ledger is silent. Touching real money or real legal stakes alone doesn't move something up a Level. Only a failure mode that can stay hidden does.
The hour ranges under each example are rough estimates. They firm up once discovery locks the actual scope.
Level One, sign-in, accounts, and CRUD: the pieces are known$13,000–26,000
Sign-in and accounts. A backend that stores and returns data. A front end onto data you already own or collect. The pieces are known; the work is composing them correctly and fast, and any mistake here is loud. It fails visibly and gets caught.
Tasks, assignments, and status boards for a small team: who's doing what, what's overdue, what's done. Sign-in, real accounts, and a backend that stores and returns the team's own data. The pieces are all known quantities. The work is composing them correctly and fast.
Built with: Next.js · Supabase (tasks, status data) · Clerk · Tailwind/shadcn
Rough estimate: ~160 hrs.
Prove it first, as a Prototype: $6,000
Standard $22,000–26,000Reduced $16,300–19,200
Proving the task and status workflow first as a Prototype, at the Prototype tier's own Level One rate, answers whether it feels right for how this specific team works before the MVP build starts. Most of the mechanic's own hours are already spent there; wiring it into real accounts and the team's own data is still real work.
A structured place for a small business to organize its own transactions against a standard chart of accounts, so the numbers are ready to hand to a bookkeeper or accountant instead of untangled from a spreadsheet first.
Built with: Next.js · Supabase (account structure) · Clerk · Claude or a comparable model, billed by usage
Rough estimate: 80–100 hrs.
Prove it first, as a Prototype: $4,500
Standard $14,000–18,000Reduced $9,500–12,200
Proving the suggested account structure first as a Prototype, at the Prototype tier's own Level One rate, answers whether it's actually useful for this specific business before the MVP build starts. Most of the mechanic's own hours are already spent there; wiring it into a real, running MVP is still real work.
Reads in a business's own CSV exports, bank and card statements, categorizes the transactions, and keeps a running, correctable record. A backend that stores and returns the client's own data, a front end onto it. Nothing here needs to guess.
Built with: Next.js · Supabase (categorized transaction history) · Clerk
Rough estimate: 70–90 hrs.
Prove it first, as a Prototype: $4,000
Standard $13,000–17,000Reduced $8,400–10,800
Proving the categorization and reporting logic first as a Prototype, matched to the Prototype tier's own CSV dashboard pattern, answers whether the output actually produces something a bookkeeper trusts before the MVP build starts. Most of the mechanic's own hours are already spent there; wiring it into the client's own accounts and data is still real work.
A place for landlords managing five to twenty units to collect rent, share lease documents, and send maintenance requests and reminders, with tenant accounts and a landlord account seeing the same real data.
Built with: Next.js · Supabase (units, tenants, rent records, documents) · Clerk (landlord and tenant roles) · Resend (rent and maintenance notifications)
Rough estimate: 100–120 hrs.
Prove it first, as a Prototype: $6,000–6,750
Standard $20,000–24,000Reduced $15,400–18,600
Proving the rent, maintenance, and document flow first as a Prototype, in the same range Prototype pricing already uses for an always-on database and real identity, answers whether it actually matches how this landlord works day to day before the MVP build starts. Most of the mechanic's own hours are already spent there; wiring it into real tenant and landlord accounts is still real work.
A tenant portal for landlords touches real rent payments and real lease documents, which made it look like it belonged higher. It doesn't. A failed rent reminder is visible and correctable. The landlord or tenant notices within days. A failed document upload is an immediate, obvious error. There's no race condition creating or destroying value in the background. Real stakes, loud failure modes: Level One.
Level Two, real-time-adjacent work, one AI doing one job well, a genuine two-sided system$18,000–38,000
More moving pieces than Level One, but the failure modes are still visible: a booking that fails to complete, an analysis that comes back wrong, a match that doesn't happen. Not silent corruption of something already real.
A small business's phone gets answered by a voice AI pipeline that books, reschedules, and cancels appointments directly against a real calendar, day and night, without anyone on staff picking up. If a booking fails to go through, that's visible immediately. Nothing gets silently double-booked or lost.
Built with: Next.js · Supabase (bookings, calendar data) · Clerk · LiveKit · a speech-to-text/text-to-speech pipeline (Deepgram or AssemblyAI, plus ElevenLabs or Google TTS) · Claude or a comparable model, billed by usage
Rough estimate: 160–200 hrs.
Prove it first, as a Prototype: $8,500
Standard $28,000–35,000Reduced $19,800–24,700
Proving the voice pipeline first as a Prototype, matched to the Prototype tier's own 'Email triage and draft replies' pattern (always-on database, real identity, the AI making a real judgment call), means the hardest part of the mechanic is already retired going in. Most of the mechanic's own hours are already spent there. What's left is wiring it into a real calendar, real accounts, and the rest of the MVP around it.
Ingests a batch of user-interview transcripts and surfaces themes, quotes, and recurring pain points for a product team, instead of someone reading every transcript by hand. If the analysis comes back wrong, that's a bad read a person catches on review, not a hidden error compounding somewhere downstream.
Built with: Next.js · Supabase (transcript storage) · Clerk · Claude or a comparable model, billed by usage
Rough estimate: 100–130 hrs.
Prove it first, as a Prototype: $8,000–9,000
Standard $18,000–23,000Reduced $13,300–17,000
Proving the analysis itself first, whether it actually pulls useful, correct themes out of real transcripts, as a Prototype answers the real question before the MVP build starts. Most of the mechanic's own hours are already spent there. What's left is building the accounts, the upload flow, and the review interface around it.
A two-sided marketplace connecting home cooks with local diners: listings on one side, discovery and ordering on the other, both sides seeing the same real, live state. A genuine two-sided system, not two independent features bolted together.
Built with: Next.js · Supabase (listings, orders, accounts) · Clerk (two roles) · Stripe
Rough estimate: 180–220 hrs.
This one doesn't get a 'prove it first' callout either. The risk is whether both sides of the marketplace actually show up and transact under real, sustained use, not a single mechanic you'd prove once in a demo, the same reasoning behind the ledger example below.
Level Three, three genuinely different reasons a system can be risky$26,000–60,000
Level Three isn't one thing scaled up. It's three genuinely different reasons a system can be expensive and risky, and knowing which one applies changes what actually gets built.
Real money or real-value assets change hands, concurrently, and a miscalculation doesn't crash. It quietly corrupts a balance or double-sells something, surfacing later as a dispute after value has already moved.
Built with: Next.js · Supabase (transaction ledger, listings) · Clerk · Stripe
Rough estimate: 250–350 hrs.
This one doesn't get a 'prove it first' callout. The risk isn't a single mechanic you'd prove once in a demo. It's whether the system holds up under real, concurrent load over time, which a prototype-scale proof can't actually test.
A live head-to-head match between two devices, built on the same shape as a shipped, bootstrapped multiplayer puzzle game with several modes: one player sends a fast clue, the other races the clock to guess it, and the round has to stay perfectly synced the whole time. Any lag, any desync, any dropped connection breaks the round in a way both players notice immediately, which is exactly the kind of risk that's provable only by running it live.
Built with: Next.js or React Native · Supabase (accounts, matchmaking, leaderboards) · a real-time layer for the live match itself (Supabase Realtime or a dedicated low-latency service) · Clerk
Rough estimate: 140–180 hrs.
Prove it first, as a Prototype: $7,750–9,000
Standard $26,000–32,000Reduced $15,800–19,600
Proving the real-time match mechanic first as a Prototype, matched to the Prototype tier's own 'Restaurant table-status board' pattern (always-on database, real identity, a live always-open connection, and genuinely hard state logic), means the sync and timing risk is already retired going in. Most of the mechanic's own hours are already spent proving it holds up under real devices and real latency. What's left is wiring it into real accounts, matchmaking, and leaderboards, which is still real work of its own.
Live spoken conversation practice with an AI built to sound like a native speaker, not read from a script. It has to listen while it's still deciding what to say, handle a learner interrupting mid-sentence the way a real conversation partner would, and adjust its own pacing for someone who's still learning the language. The speech-to-text underneath has to hold up against accented, imperfect learner speech, which is a harder recognition problem than a native speaker calling a business to book an appointment. That difference is what actually separates this from the receptionist example above.
Built with: Next.js · Supabase (accounts, session history) · Clerk · LiveKit · speech-to-text tuned for accented speech · a text-to-speech provider · Claude or a comparable model, billed by usage
Rough estimate: 220–280 hrs.
Prove it first, as a Prototype: $10,500–12,000
Standard $38,000–46,000Reduced $23,100–28,100
The voice pipeline and adaptive-conversation judgment are the hardest, least-certain part of this build. Proving it first as a Prototype, at the top of the Prototype tier's own range, retires the biggest unknown before the MVP build starts. Most of the mechanic's own hours are already spent there, but the specialist's polish work and the integration into the MVP's accounts and lesson flow still carry real time, which is why the reduction is smaller in proportion than the game example above.
After the warranty window
The ninety days aren't a cliff. If you want us back in the room, that's the named advisory option: a block of our time, on your terms, not a subscription running quietly in the background. If you'd rather take it from here, everything you're handed (the code, the architecture, the decisions behind it) is yours outright the day it ships.
The foundation everything else sits on, and what it buys you the first time you need to change your mind.
- The layered architectureIncluded
- Your app is built in three separate layers: what a person sees, the logic that decides what happens, and the data underneath it. Because they're kept apart instead of tangled together, you can redesign a screen, rework a flow, or run an A/B test on a whole feature without touching the layers underneath it. That includes building the frontend, setting up the databases underneath it, and wiring up communication with the third-party services your product depends on. Design is genuinely part of the scope too, not an afterthought, and building it well takes real time, which is part of why the hours on a Level look the way they do.This is about the shape of what you're handed, not an ongoing promise to keep rebuilding it for you. If you want us doing that work regularly after launch, that's the advisory option below, named and priced on its own terms.What this means
- Who owns itIncluded
- The code, the architecture, the decisions behind it: yours, outright, the day it ships. Nothing about this build depends on you staying a client to keep using what you paid for.What this means
The build ends with an actual handoff, not a login you inherit and a channel that goes quiet. Four pieces, always included.
- DocumentationIncluded
- Auto-generated component and API references that stay current as the system is built, plus a written layer on top: an architecture overview, a setup guide, and a log of the decisions that shaped it and why.What this means
- Your project historyIncluded
- The full build record, every ticket and every change, handed to you in a portable file you can open and keep. Not a login to a workspace you'll lose access to eventually. Something that's actually yours.What this means
- Ninety-day warrantyIncluded
- For ninety days after launch, if something doesn't do what the spec said it would do, we fix it. That's the whole scope. It doesn't cover new features you think of afterward, and it doesn't cover the market deciding it doesn't want what you asked for. It covers whether we built what we agreed to build.What this means
- Advisory, if you want itScoped with you
- A named, paid, bounded option for after the warranty window closes: a block of our time, on your terms, whenever you actually need us back in the room. It isn't a retainer, and we won't try to talk you into one. Our aim is to keep building new MVPs for new founders, not to become anyone's permanent outsourced dev team.What this means
- Privacy and compliance basicsIncluded
- Bundled in by default: cookie consent, encryption, a real authentication provider, and a real path for a user to delete their own data. We architect it. You own it, and the legal and policy decisions on top of it stay yours.What this means
The services your MVP actually runs on. Named upfront and billed to you directly, because it's an ongoing cost of keeping the thing alive, not a one-time cost of building it.
- The default bundle$100–150/mo
- Hosting to keep it live. A database, authentication, and file storage, set up twice, once for development and once for the real thing, so testing never touches real user data. A dedicated authentication provider. Transactional email. Error tracking, so we know something broke before your users have to tell you. Analytics, session replay, and feature flags, so you can actually see what people are doing with it.What this means
- Why a full stack instead of the cheapest option
- A bare database is cheaper sitting idle. It also leaves you, or us, to assemble authentication, file storage, and access rules by hand. We chose a stack that comes with the whole thing working together, because you're a founder trying to learn whether people want your product, not a founder who also wants to be a part time database administrator.What this means
- Going deeper, if you want itScoped with you
- Usability testing and a real feedback channel for the people already using it. Optional, your call, billed separately if you want it added.What this means
A separate menu from the bundle above, identified and estimated during discovery because what you actually need here depends entirely on what you're building. We estimate it and we cap it. We don't control it and we don't mark it up.
- AI and language model callsScoped with you
- If your product calls an AI model to do its thinking, those calls are billed by usage. We estimate the range during discovery and keep a cap on it.What this means
- Vector search and retrievalScoped with you
- If your product needs to search across its own content or remember something about a user over time, that's a vector database, scoped and priced as its own line.What this means
- Voice and audio pipelinesScoped with you
- If your product listens or speaks: real time voice infrastructure plus speech to text and text to speech, the same category of stack running underneath Discourser.ai, the voice practice tool we built ourselves. Estimated up front, not discovered on your first bill.What this means
- Image and media generationScoped with you
- Generated imagery or media, billed by usage from whichever provider fits what you're building.What this means
- PaymentsScoped with you
- Processing fees from a provider like Stripe. Standard, and outside anyone's control but theirs.What this means
Beyond this tier
- Genuine PHI / HIPAA-custodial work
- There is no official U.S. government HIPAA certification. Compliance is entirely self-attested, and even a paid third-party readiness assessment ($5,000–40,000+ for a small practice, consultants at $250–350/hr) buys a private opinion, not a government-backed guarantee. That liability has no external backstop, and it belongs with a specialist who carries real insurance and legal counsel for exactly this exposure, not with a solo studio. We'll say so directly rather than take it on.
- Work we'd send you elsewhere for
- Hardware and physical production. Advanced animated or game production, real 3D and motion work at a studio's depth. Deep machine learning work: training a model from scratch, or building and tuning a vector database rather than composing an existing one. We build the frontend and the connective layer around systems like these, not the underlying engine itself. If it's useful, that can mean a Node-based service layer that calls into a specialized system someone else runs, wired cleanly into your MVP, without us building that system ourselves.
- Not a standing dev team
- We're not a dev shop, and we're not your ongoing offshore support team. What you get is a named, bounded engagement with a real handoff at the end of it, not standing outsourced development capacity.