BYOB Credit-Based Pricing Explained: No Subscriptions, Pay As You Go

BYOB credit-based pricing, no subscriptions, pay as you go #

A credit is a prepaid token for AI work. You buy credits once, then spend them when the machine thinks on your behalf. Generation, refactors, vision, voice, deployment. That is the full list of what costs money. Everything else, editing by hand, browsing, version history, local testing, exporting code, is free.

Think of it as a prepaid electricity meter for your builds. The meter spins only while the AI is working. When you stop, the meter stops, and whatever balance remains sits there waiting. Purchased credits never expire, as stated in BYOB's own FAQ on byob.studio. No monthly bill arrives for months you barely built anything.

Why do credits exist at all? #

Every AI operation burns real tokens on someone's GPU, and tokens have wholesale prices that differ wildly. OpenAI's model catalog lists flagship input at $10 per million tokens next to budget models at $0.20, a fiftyfold spread for the same unit. Output tokens cost more still. A platform that charges a flat monthly fee has to average all of that into one number and hope heavy users do not eat the margin. A credit system skips the averaging. You pay for what the machine actually consumed.

This is not a BYOB invention. Stripe documents the same pattern under a different name in their usage-based billing docs: pay as you go, fixed fee with overage, and credit burndown, where customers prepay and draw down a balance as they consume. BYOB runs the third model. You prepay, then burn down.

How do credits work? #

Credits are BYOB's internal currency. You buy credits with real money, then spend credits when you use AI features.

What consumes credits:

• AI code generation, building features, pages, components
• AI refactoring and edits, modifying existing code through chat
• Vision processing, converting screenshots or designs to code
• Voice commands, building with voice input
• One-click deployments, publishing your site live

What stays free:

• Manual code editing in the editor
• Browsing your project
• Version control, saving and restoring snapshots
• Viewing history
• Testing your site locally
• Exporting code

You pay for AI assistance and infrastructure. Everything else costs nothing. That split is deliberate. The platform wants you living inside your project, poking at it, reading history, trying things by hand. It charges only at the moments where metered compute runs on your behalf.

How is pricing set by region? #

BYOB detects your location and shows appropriate pricing, as listed on byob.studio.

India pays ₹1 per credit, with UPI, cards, and local payment methods. International pays $1 per 100 credits, effectively $0.01 per credit, through international cards. The gap reflects purchasing power differences, and region detection happens automatically with no manual selection.

Starter accounts also include free credits to try the platform, and paid tiers exist alongside top-ups: Pro at $25 per month with 2,000 plus credits, Max at $79 with 8,000 plus credits, per the pricing section on byob.studio. Credits and subscriptions coexist. Top up when you explore, subscribe when usage turns steady.

How much do things actually cost? #

Try it: BYOB credit estimator

Real costs for common operations, at BYOB's stated rates:

Building a landing page #

Prompt: "Build a SaaS landing page with hero, features, pricing, and contact form"

The AI generates a hero section, a three-column features block, a three-tier pricing table, a validated contact form, and responsive styles throughout. Expect roughly 50 to 80 credits, which works out near ₹50 to ₹80 or $0.50 to $0.80.

Adding a blog to an existing site #

Prompt: "Add a blog section with post list, individual post pages, and categories"

This touches listing pages, dynamic routes, a category system, and a posts schema. Expect roughly 60 to 100 credits, near ₹60 to ₹100 or $0.60 to $1.00. Multi-file changes cost more than single-file ones because more tokens move in both directions.

Refactoring an existing component #

Prompt: "Make the hero background a gradient animation instead of static"

Small scoped edits like this run roughly 15 to 25 credits. The model reads the file, rewrites the slice, and returns. Cheap, because context stays narrow.

Deployment #

Clicking Publish compiles the SvelteKit project, bundles and optimizes assets, pushes to the edge network, and provisions SSL. Expect around 5 credits, near ₹5 or $0.05. Deployments are the cheapest line item on the menu, which is exactly right. You should deploy often without doing mental math.

As a rough ladder: simple pages cost 50 to 100 credits, complex features cost 100 to 200, small edits cost 10 to 30, deployments cost about 5.

What does typical monthly usage look like? #

Light builders, one or two small projects a month, generating a few pages with a dozen edits and a handful of deploys, land near 300 credits. That is roughly ₹300 or $3.00.

Active builders, multiple projects with ongoing iteration, dozens of edits and twenty deploys, land near 1,000 credits. Roughly ₹1,000 or $10.00.

Heavy builders, agencies shipping client sites, dozens of pages and a hundred plus edits with fifty deploys, land near 3,000 credits. Roughly ₹3,000 or $30.00.

Most solo builders sit between light and active. Agencies average near ₹2,000 to ₹5,000 monthly. These are planning numbers, not promises. Your prompt discipline moves them more than any pricing page does.

How does credit pricing compare with subscription pricing? #

Subscription tools charge $20 to $80 monthly whether you touch them or not, with annual commitments unlocking the better rates. BYOB credits charge only for actual operations, never expire once purchased, and require no commitment.

Credits win for bursty schedules. Build two projects in March, nothing until July, and you paid for March twice over nothing. Subscriptions win for constant daily generation, where unlimited use inside fair use limits beats per-operation metering.

The break-even sits near ₹2,000 ($20) in monthly credit burn. Consistently above that, compare against Pro or Max. Consistently below, credits save money. Stripe's own docs describe the hybrid the industry converged on, a flat fee covering baseline plus metered overage above it, in their usage-based billing guide. BYOB mirrors that shape: a subscription tier for the steady middle, top-ups for the spikes.

What are free starter credits? #

New accounts receive 100 starter credits, worth about ₹100 or $1.00. That covers one or two simple landing pages, a round of AI edits, and three to five deployments. Enough to feel the whole loop.

Starter credits behave exactly like paid credits with no restrictions. They expire after 30 days if unused, which exists to prevent abuse, not to pressure you. Purchased credits carry no expiry at all.

How does the referral program work? #

BYOB pays both sides of an introduction. When someone signs up through your link, you get 5 credits. When they make their first purchase, you get 10 more and they get 10 bonus credits. Fifteen credits per converted referral buys roughly three deployments or a solid page edit, free.

Find your link behind the Invite button in the dashboard header or on your profile page, in the format https://byob.studio/auth/signin?ref=YOUR_UNIQUE_CODE. There is no cap. Share it in communities, tutorials, videos, or one direct message to a founder friend who keeps saying they need a site.

How to purchase credits #

Click Buy Credits in the interface, choose an amount starting at a 100 credit minimum, and pay by card or UPI. Credits land instantly. Pricing per credit stays flat across volumes today, with no bulk discounts, though high-volume agencies can ask about custom plans.

Payments run through Paddle, which means checkout behavior follows Paddle's documented patterns: sandboxed test mode while integrating, live keys for real money, and client-side tokens that are safe to publish in frontend code, as described in Paddle's inline checkout guide. Webhook receipts get verified by signature before any entitlement changes, the standard Paddle flow documented in their signature verification guide. You never see this machinery. It matters only because money handling borrows its reliability from infrastructure built for exactly that.

What are the credit expiration and refund rules? #

Purchased credits never expire. Buy 1,000 today, spend them across two years, feel nothing. Free and promotional credits expire after 30 days. Credits are non-refundable once purchased, which is why the starter credits exist. Evaluate risk-free first, then pay.

How does credit usage tracking work? #

BYOB keeps the balance visible in the interface and logs every transaction: minus 50 for a generated hero, minus 5 for a deploy, plus 100 for a purchase, each timestamped. Read that log monthly. It tells you whether your burn comes from ambitious builds, which is good, or from reruns of vague prompts, which is fixable.

How do you optimize credit usage? #

Detailed prompts cost about the same as vague ones but succeed on the first try far more often. "Build a hero with gradient, CTA button, and benefits list" and "build a hero" burn similar tokens, yet the first rarely needs a second pass. Specificity is the cheapest optimization available.

A few more habits that compound:

• Edit tiny changes by hand. Button color tweaks do not need a model.
• Test locally before deploying. One working deploy beats three hopeful ones.
• Batch related changes into one coherent instruction instead of three separate asks.
• Keep stable instructions stable. Anthropic's prompt caching docs show that unchanged prefixes can be reused across calls at roughly a tenth of base input price, while rewritten intent forces full reprocessing. Stop restating the design system every turn.

Transparency and no hidden costs #

All costs run through credits. Generation, global CDN hosting, SSL certificates, bandwidth within fair use, version storage, support. Export your code anytime with no lock-in fee. The price you see before an AI run is the price you pay.

What about enterprise and high-volume users? #

Flat per-credit pricing holds for everyone today. Volume discounts for large prepaid blocks and custom agency plans sit on the roadmap. If your shop ships dozens of client sites monthly, talk to support instead of guessing.

What is the token math behind a page? #

Credits abstract tokens, but seeing one layer down helps you spend wisely. A landing page prompt might send a few thousand input tokens of instructions plus tens of thousands of project context, and receive ten or twenty thousand output tokens of code. At wholesale rates from OpenAI's model catalog, flagship input near $10 per million tokens and output near $50, a single big generation costs the provider cents, not dollars. BYOB wraps that wholesale cost plus hosting, support, and margin into the 50 to 80 credit retail price.

Two levers shrink that cost per outcome. First, model routing: budget models near $0.20 per million input tokens handle routine work at a fiftieth of flagship rates, which is why BYOB supports multiple models with per-session lock. Second, caching: unchanged prompt prefixes get reused across calls at roughly a tenth of base input price, per Anthropic's prompt caching docs. Stable instructions plus the right model means your credits buy more shipped features per dollar. None of this requires your attention day to day. It runs underneath, but knowing it exists explains why tight prompts and steady context outperform rambling threads economically.

How do you choose between top-ups, Pro, and Max? #

Three ways to pay, matched to three usage shapes.

Top-ups fit explorers. Building your first project, testing the platform, shipping one site per quarter. Buy 100 credits at a time, spend across months, never think about billing dates. The per-credit rate stays flat, so small buyers face no penalty.

Pro at $25 monthly with 2,000 plus included credits fits regular builders. One active project with weekly iteration, a freelancer with two concurrent clients, a founder iterating toward launch. The subscription covers the steady middle while top-ups absorb spikes, the hybrid Stripe documents as flat fee plus overage in their usage-based billing guide.

Max at $79 monthly with 8,000 plus credits fits production shops. Agencies with parallel sprints, studios shipping weekly, teams where BYOB is daily infrastructure. At that volume the included credits beat equivalent top-ups, and the predictability simplifies client billing: one subscription line item, spread across projects.

The migration path runs upward naturally. Start on top-ups, watch two months of burn, graduate when the math says so. Downgrades work the same direction. No contracts trap you at a tier your workload outgrew or never reached.

Teams, clients, and passing costs through #

Agencies ask the margin question immediately: if a client site burns 400 credits, about $4 at international rates, and the project bills $3,000, the AI cost rounds to zero. It vanishes inside the project fee. Track it anyway. Per-project credit logs make cost-of-goods-sold visible per client, which matters when one client's endless revisions burn ten times the average. Value pricing plus measured costs equals margins you can defend.

Freelancers on retainers get a second benefit. Monthly maintenance, headline swaps, seasonal banners, new testimonials, costs single-digit credits per change and deploys for about five. A $200 monthly retainer against maybe $2 of credit burn is the kind of ratio that funds vacations. Quote the retainer on response time and reliability, not on token counts clients will never understand.

Which habits cost nothing? #

The cheapest credits are the ones you never spend. A surprising amount of professional workflow sits entirely in the free zone.

Read version history before regenerating. The answer you need often exists three snapshots back, restorable free instead of rebuildable for fifty credits. Edit copy by hand in the split-view editor. Headlines, prices, bios, and button labels need no model involvement. Test locally until the work is right, then deploy once instead of publishing drafts to production for inspection. Preview URLs exist for exactly this.

Batch review feedback before prompting. Five client comments collected across a week become one coherent instruction instead of five separate generations with overlapping context. The discipline mirrors good engineering everywhere: fewer, larger, well-specified operations beat scattered micro-asks on cost and on quality.

What counts as one operation in metering edge cases? #

Metering questions always arrive eventually, so here are the answers upfront. A generation that retries after a model error counts once for the successful output, not per attempt, because you should not pay for the platform's stumbles. A conversation turn that produces no code, like asking a question about your project, consumes less than a generation but still processes tokens, so it meters lightly rather than zero. Vision processing meters by image analysis depth: a quick "match this screenshot" costs less than a full design-to-code conversion with multiple views.

Voice commands meter like their text equivalents plus transcription. A spoken paragraph becomes text, then runs the same generation path. Deployments meter per publish event, so mashing the button five times in a minute spends five times. Preview refreshes during development cost nothing. The line stays consistent throughout: metered compute costs credits, your own time and attention cost nothing.

When a charge looks wrong, the usage log settles it. Every entry carries timestamp, operation, and amount. Compare the entry against the list above before assuming a bug. Genuine metering errors get corrected by support. Perceived errors, nine times out of ten, trace to an operation the user forgot they ran, like the overnight deploy from a teammate in another timezone.

What are promotional balances and credit gifts? #

Starter credits, referral earnings, and occasional promotional grants all land in one balance and spend identically to purchased credits. The only distinction is expiry: free balances carry the 30-day window while purchased balances never expire. Spend free credits first on experiments and learning, since their clock ticks, and save purchased credits for production work that must never stall mid-sprint. The usage log does not segregate sources, so track large referral months mentally when reconciling burn against expectations.

What are the trade-offs? #

Credits fit bursts: buy once, spend when the model works, purchased balances never expire, free starter 100 expire after 30 days.

Pick Pro or Max bundles when daily output is steady and predictability beats per operation control. Pick top-ups when builds come in bursts and idle months should cost nothing.

What we learned building this #

Credits in BYOB map to metered compute plus hosting, with purchased credits that never expire and a balance visible in the app, logic mirrored in the plan tiers and the credit purchase dialog. We show estimates before each generation and log exact cost after, so tight prompts and cached prefixes stretch the same balance. We verified the credit estimator at https://byob.studio/tools/byob-credit-estimator returns 200 and point new builders there first.

Who this is for (and who should skip it) #

This guide helps if you build in bursts and want to pay for AI operations only when the model works.

If you generate daily at high volume and want flat predictability, compare the Pro and Max bundles instead of pure top ups.

• Best for beginners building in bursts and paying per use.
• Best for small business owners testing ideas without a monthly plan.
• Best for freelancers matching spend to active client work.

Frequently asked questions #

What happens if I run out of credits? #

AI features and deployments pause until you top up. Existing sites stay online and manual editing keeps working. Nothing breaks, nothing disappears.

Can I get a refund if I don't like BYOB? #

Credits are non-refundable. The 100 starter credits exist so you can judge the platform before spending anything.

Do credits cover hosting costs? #

Yes. Hosting, CDN, SSL, and bandwidth sit inside credit pricing. Live sites carry no ongoing drain while they serve traffic normally.

How do I know how many credits something will use? #

BYOB shows an estimate before running AI operations, and actual cost may vary slightly with complexity. The usage log then shows exactly what each run consumed.

Can I share credits with team members? #

Not currently. Each account holds its own balance. Shared pools may arrive with team features.

Start building with free credits. Try BYOB now →