--- slug: cost-to-vibe-code-an-app title: "How Much Does It Actually Cost to Vibe-Code an App?" excerpt: "The real cost to build an app with AI in 2026: tooling, infra, launch fees. An itemised receipt for shipping a real weekend project." primaryKeyword: "cost to build an app with AI" publishedAt: 2026-05-10 readingTimeMin: 8 author: "Robert Boylan" tags: - indie-dev - vibe-coding - pricing - tool-comparison - ai-coding-tools --- Someone shipped a small habit-tracker last weekend. Two days of building, one Sunday evening of polish, a working app live on the internet. A friend at brunch asked the obvious question: "what did that cost you?" They froze. Not because the answer was embarrassing, but because they hadn't actually added it up. There was the Lovable subscription. The domain. That month they upgraded ChatGPT to think through the data model. The Vercel bill that was zero, until the third week of traffic when it wasn't. The Sentry account they signed up for after the first real bug. This post is the itemised receipt none of those marketing pages give you. The honest cost to build an app with AI in 2026: what the tooling actually runs, what the infrastructure costs once you stop being on the free tier, and what the worked example of a real indie weekend build looks like three months in. ## Why "free" is the wrong answer to start with You'll see headlines that say you can build an app for nothing. Technically, that's true for about a week. Most AI coding tools have a free tier. Most hosting platforms have a free tier. Most databases have a free tier. If you stack the free tiers right and your app gets four users, you can absolutely run it for zero. The honest version: the moment your project has a custom domain, more than a handful of users, or any kind of paid feature, the bill starts. Not a big bill. But a real one, and one most "build an app with AI" guides skip past. So instead of telling you it's free, we're going to walk through what the receipt actually looks like. Tooling, infra, launch costs, and the real example at the end. The numbers below are in euros, with US dollars in parens where it helps. Prices move, especially the AI tooling ones. Read these as "early 2026, ballpark" rather than gospel. ## The tooling line: what writes the code This is where most of the early money goes. You're paying for AI subscriptions, and the question isn't really "do I need one" but "which two or three do I need." A typical indie tooling stack: - **Lovable** (a browser-based AI app builder you describe ideas to). About €23/month ($25) for the starter paid tier. Free tier exists but the credit limit kicks in fast. - **Cursor** (an AI-native code editor, basically VS Code with AI built in). €18/month ($20) for Pro. Free tier is genuinely usable for very light work. - **Claude Code** (a command-line AI agent that reads and edits whole codebases). Usage-based. Light use is €5-10/month, a heavy weekend can run €40-60. - **v0** (a design-focused AI tool from Vercel that turns prompts into React UIs). €18/month ($20) for the paid tier. - **ChatGPT Plus** (the paid OpenAI consumer tier). €19/month ($20). Useful for the planning and thinking-out-loud work, less so for production code generation. You almost never use all of these. A common pairing is one builder-style tool plus one developer-style tool. Lovable plus Cursor, or v0 plus Claude Code. That puts your tooling line at roughly €40-50/month if you stay paid on both. If you go all-in on AI usage during a build sprint, the realistic ceiling is closer to €80/month. If you're just dabbling on weekends, you can get away with €20. We've covered [which AI coding tool fits your situation](/blog/which-ai-coding-tool-should-you-use) in more depth if the choice itself feels paralysing. The short version: pick one builder, pick one editor, stop reading comparison threads. ## The infra line: what runs the app Now your app exists. It needs to live somewhere. This is where the bill gets sneakier, because most of these have generous free tiers right up until they don't. The typical infra stack for an indie app: - **Vercel** (a hosting service for web apps, especially ones built with Next.js or React). Free tier handles a lot of traffic. The Pro tier kicks in at €18/month ($20) once you cross bandwidth limits, want password-protected previews, or need team features. - **Supabase** (an all-in-one backend with a Postgres database, auth, and file storage). Free tier covers small projects. Pro is €23/month ($25), and you'll usually hit it once your database goes past 500MB or you need daily backups. - **A custom domain**. €10-15/year on Namecheap, Cloudflare, or Porkbun. Skip the .com if it's taken; the .app and .io options are fine and often cheaper. - **Resend or Brevo** (transactional email services that send things like signup confirmations and password resets). Free tier on Resend gives you 3,000 emails/month, which is more than most weekend apps will ever send. Brevo similarly has a free tier with 300 emails/day. The honest pattern: month one is free. Month two is usually free. Month three is when one of these tips into a paid tier, usually Supabase first because the database fills up, or Vercel if your app gets a real traffic spike. Realistic infra spend at three months in, for a small but live app: €25-45/month. Drop one of those (live without daily backups, or stay on the free email tier) and you can keep it under €20. ## The launch line: what comes online when you go live These are the costs nobody warns you about, because they're optional right up until the morning your app actually gets used. The launch-time list: - **Stripe** (a payment processor). No monthly fee, but takes 1.5% + €0.25 per EU transaction (2.9% + $0.30 in the US). Worth saying out loud: if you charge €10 and the customer pays with a US card, you keep about €9.40. Plan your prices accordingly. We've written about [how to price the app once it exists](/blog/pricing-your-indie-app) if you're staring at the pricing page wondering what to put. - **Plausible or PostHog** (analytics tools that show you what visitors are actually doing). Plausible is €9/month ($10) for the starter tier, privacy-first and simple. PostHog has a generous free tier and gets pricey if you cross the event limit. Either is fine; pick one. - **Sentry** (an error-tracking service that pings you when the app blows up). Free tier covers most indie apps. Paid kicks in at €26/month ($28) once you cross the error volume limit. - **A support email or helpdesk**. Most people just use Gmail until they can't. Crisp or Help Scout sit at €25-30/month if you grow into wanting a real inbox tool. Realistic launch-line spend: €10-30/month, depending on which of these you actually need. Most weekend builders skip Sentry until the first user-reported bug, skip the helpdesk forever, and pick one analytics tool. ## A worked example: three months of an indie weekend build Let's put a real number on this. Imagine the habit-tracker from the opener: built over a weekend, launched the next Friday, three months later it has 80 signups and seven paid users. **Upfront, the first month:** - Lovable Starter: €23 - Cursor Pro: €18 (signed up halfway through to clean up the AI-generated code) - Custom domain: €12 for the year - Supabase: €0 (free tier) - Vercel: €0 (free tier) - Resend: €0 (free tier) - Stripe: €0 to set up - Plausible: €9 **Month one total: roughly €62, plus the €12 domain that's a one-off.** **Month three, after some growth:** - Lovable Starter: €23 - Cursor Pro: €18 - Claude Code (used for a refactor sprint): €15 - Supabase Pro: €23 (database crossed 500MB) - Vercel: €0 (still on free) - Resend: €0 - Plausible: €9 - Sentry: €0 (still under free tier) **Month three total: roughly €88.** So three months in, the running total is somewhere around €240. Realised revenue from those seven paid users at €5/month is roughly €70. The app isn't profitable yet, but the gap is small enough to close in a few months if growth holds. The takeaway from this receipt: the upfront cost is genuinely low (€60-100 to get going), but the steady monthly run rate creeps up faster than people expect. Budget €25-50/month as your real ongoing cost once the app is live and growing. ## The cost nobody puts on the receipt The biggest expense building an app with AI isn't on the bill. It's the time you spend re-prompting because you didn't think the thing through. Every wasted Cursor round costs a little. Every Claude Code session that goes sideways because the description was vague costs more, because Claude Code doesn't waste one line of code, it wastes fifteen across six files. A single hour of "I should have planned this first" easily burns €5-10 of AI credits and an evening of momentum. The math here is unkind. If you spend ten hours building and four of them are "this isn't what I meant" cycles, your effective cost per shipped feature is roughly double what it should be. The tooling line in the receipt above assumes you're roughly on track. If you're prompting blind, double it. The fix isn't more prompt engineering tricks. It's writing down what you're actually building, on what stack, for whom, and what the day-one feature set is, before you open a single AI tool. We've covered [picking a tech stack without overthinking it](/blog/how-to-choose-a-tech-stack) if that's the part that stalls you. Most people skip the planning step because it feels boring next to the "watch the AI build it" part. That's exactly why their bill ends up bigger than it needed to be. ## So what does this actually cost? The honest answer to "how much does it cost to build an app with AI" is: about €60-100 to start, and €25-50/month to keep running once you're real. Plus your time, plus the wasted-credits tax if you start prompting before you've thought. The tooling and the infra are cheaper than they've ever been. The thinking is the same as it always was. Planning before you start prompting is what cuts the biggest hidden cost on the receipt: the AI tokens and re-rolls you spent figuring out what you wanted while you were already building it. That's the gap Draftlytic is built for. You describe the idea, it produces a structured spec ready to drop into Lovable, Cursor, Claude Code, or whatever you're using, and the bill at the end of the month looks a lot more like the optimistic version of the receipt above.