Lovable Pricing In 2026: An Honest Look At The Credit Casino.

Every few months Lovable rearranges its pricing furniture, and every few months the indie-hacker internet has the same argument about whether the new layout is generous, predatory, or merely confusing. The 2026 version is, in fairness, the cleanest the platform has shipped — one credit balance, one dashboard, one set of numbers to lose sleep over. It is also, like every credit-based AI product, a system designed by people whose margin depends on you not quite understanding what a credit is. So let us go through it slowly, with a calculator and the appropriate level of suspicion.

The plans, in plain English

There are four tiers. Free, Pro, Business and Enterprise. Free costs nothing and gives you five build credits a day, capped at thirty per calendar month — enough to kick the tyres, build a landing page, and decide whether you like the editor. Pro starts at twenty-five dollars a month for a hundred credits and scales up in steps to ten thousand credits at two thousand two hundred and fifty a month. Business is the same credit tiers at roughly double the price, and what you are paying double for is SSO, a data-training opt-out, per-member credit limits, design templates, restricted projects and a security centre. Enterprise is the usual 'call us' arrangement, priced by how desperately your procurement team wants a contract.

The headline number most people quote — twenty-five dollars, a hundred credits — is also the worst-value tier per credit, at roughly twenty-five cents each. The price per credit slides down as the volume goes up, bottoming out somewhere around twenty-two and a half cents at the five-thousand-credit-and-above tiers. Annual billing knocks about sixteen per cent off either way. Top-ups, when you inevitably blow through your monthly allocation halfway through a Sunday afternoon, are sold at a twenty per cent premium on Pro and a hundred per cent premium on Business — fifteen dollars per fifty credits for Pro, thirty dollars per fifty for Business. The minimum top-up is fifty credits, the maximum is a thousand, and they expire in twelve months in case you were planning to hoard them like baked beans.

The bit nobody mentions: the daily credit bonus stacks

Every paid plan, in addition to its monthly allowance, gets five build credits a day with no monthly cap. That is up to a hundred and fifty extra credits a month sitting on top of your hundred-credit Pro subscription — meaning a twenty-five-dollar Pro plan can, in a busy month, deliver up to two hundred and fifty effective credits. Almost nobody factors this into their 'is Lovable expensive' tweets. It is the single most under-quoted number in the entire pricing model, and it materially changes how the platform compares to Bolt, v0 and the rest. The daily credits do not roll over and they are not transferable, but if you are building most days they quietly subsidise about a third of your real usage.

Plan mode, build mode, and the difference between a question and a bill

There are two ways to talk to the agent. Plan mode is the conversation — the 'let us discuss this before we commit to anything' mode — and it costs a flat one credit per message regardless of how clever the answer is. Build mode is the doing — the mode that actually edits your code, runs migrations, generates components and reads your repo — and since the great pricing reshuffle of July 2025 it has been billed by complexity rather than per-message. A trivial CSS tweak might be half a credit. Removing a footer might be 0.9. Adding authentication, around 1.2. A full landing page with hero images and theme work, two credits or more. These numbers are illustrative, not contractual, and that is half the problem.

Because here is the unromantic truth, and it is the single most consistent complaint across G2, Reddit and every blog post written by anyone who has used the platform for longer than a fortnight: you cannot see the cost of a build prompt before you send it. You can only see it afterwards, in a tooltip under the response. If the agent decides your innocent-looking request requires reading half your codebase, browsing the web, generating three images and writing a migration, it will quietly do all of that and present you with the bill on the way out. Roughly six in ten negative reviews mention this opacity by name. It is fixable. It has not been fixed.

There is one mercy. Plan mode is one credit a message, full stop, no matter how long the answer. Use it. If you cannot tell whether a feature will take two credits or twenty, ask in plan mode first. It is the cheapest piece of advice in this article.

The Cloud merger: one balance to rule them all

Until earlier this year, Lovable ran three parallel wallets. There were build credits, for talking to the editor. There was a Cloud balance, denominated in dollars, for hosting, database and bandwidth. And there was an AI balance, also in dollars, for the AI Gateway calls your deployed app made on behalf of its users. Three balances. Three top-ups. Three different things to forget to monitor. It was, charitably, a mess.

Some time around March of 2026 — the rollout has been gradual, and some workspaces are still on the old system — Lovable merged the three into a single credit bank. Cloud and AI dollar balances were converted into credits at the plan's rate and dumped into your general pool. The dollar-denominated auto top-ups were retired. Everything now lives in one place under Plans And Credit Usage, and one number tells you how close you are to running out of everything simultaneously. Free monthly Cloud and AI grants still exist as separate buckets that get drained first — twenty Cloud credits a month on every plan, plus four AI credits a month on the Free tier — but they all top up the same single pool when they are gone.

The good news: the underlying costs did not change. A megabyte of bandwidth costs what it always did. A Gemini call costs what it always did. The bad news: build, hosting and run-time AI now compete for the same wallet. A traffic spike on a successful app can quietly eat the credits you were saving for next week's features. Before the merge, you noticed because the dedicated Cloud balance went red. After the merge, you notice because the agent suddenly tells you to top up before it will let you ship a button colour change. This is the trade-off of unified billing across every category of SaaS: less mental overhead, less granular control. Lovable did not invent the problem. They just signed up for it.

How it stacks up against the competition

The vibe-coding pricing landscape in mid-2026 is a polite cartel of similarly-priced products with subtly different value propositions. Here are the entry tiers, with the marketing stripped off.

• →Lovable Pro — twenty-five dollars a month for a hundred credits, plus up to a hundred and fifty bonus daily credits, plus hosting, Supabase backend, AI Gateway, custom domains and a real editor in the browser.
• →Bolt.new Pro — twenty-five dollars a month for ten million tokens, no daily limit, hosting via Netlify rather than native, full-stack React in a StackBlitz WebContainer.
• →v0 Team — thirty dollars per user per month for thirty dollars of token-metered credits, plus two dollars of daily credits per user, deploys straight to Vercel, brilliant at Next.js UI components and not much else.
• →Replit Core — twenty-five dollars a month (twenty annual) for twenty-five dollars of included credits, fifty-plus languages, a real backend environment, fully hosted.
• →Cursor Pro — twenty dollars a month, no hosting at all, two usage pools per cycle, frontier models, designed for developers who already have an editor opinion.
• →Windsurf Pro — twenty dollars a month, no hosting, flat-quota model with overage at API pass-through cost, recently acquired by OpenAI which is a story for another article.
• →Base44 Starter — sixteen dollars a month annual, a hundred message credits, hosting included, owned by Wix, simpler than Lovable, less powerful, the cheapest entry ticket in the category.

The honest comparison: Lovable is the most expensive entry-tier product in the category that includes everything — editor, full-stack code generation, hosting, database, auth and an AI Gateway for the deployed app — in a single subscription. Bolt is at the same price but punts hosting to Netlify. Replit matches the price and the hosting but is aimed at a different user, the developer who wants a real terminal more than a no-code-friendly chat. Cursor and Windsurf are cheaper because they do not host anything; they are tools for people who already know how to deploy. Base44 is cheaper still but obviously the budget option in a Wix-owned future. v0 is in its own world, expensive per credit, narrow in scope, and the right answer if your entire product is a beautiful Next.js front end.

If the question is 'which one is cheapest', the answer is Base44, and you should stop reading. If the question is 'which one lets a competent operator ship a full-stack product without owning DevOps', Lovable, Replit and Bolt are within shouting distance of each other and the deciding factor is taste, stack preference and whether you can stomach Netlify's billing emails.

Hosting on Lovable: the convenience tax

Lovable Cloud is, underneath the marketing, a Supabase project plus a Cloudflare Workers edge runtime, all stitched together so neatly that you can forget there are two products in there. For the first ninety per cent of projects this is wonderful. You ask for auth, you get auth. You ask for a database table, it appears. You ask for an email when a user signs up, an Edge Function shows up to send it. There is no Dockerfile to write, no CI pipeline to configure, no SSL certificate to renew, no Render account to forget the password to. On the Free plan you get a lovable.app subdomain. On Pro and above you can point your own domain at it, and the SSL is automatic. For a one-person business or a side project, this is genuinely the best deal in the category.

The price of that convenience is the price of all convenience: you are renting. Your hosting bill is denominated in credits, which means it is denominated in 'Lovable's pricing roadmap for next quarter'. Your database lives in an instance whose size and region you do not fully control. Multi-region is not a knob you can turn. GDPR data-residency choices are limited to what the platform supports. If your app suddenly catches on, the Cloud portion of your credit burn will scale with traffic in a way that is opaque enough that you will see complaints about hundred-dollar months from people whose marketing page is 'twenty-five dollars per month'. None of this is dishonest. It is just usage-based hosting, billed in a unit invented by your landlord.

Leaving the platform: it is easier than you think

Here is the thing the pricing-page footnotes do not emphasise enough: what Lovable generates is not a proprietary file format. It is a standard TanStack Start project — React, TypeScript, Vite, the same npm ecosystem the rest of the web runs on — wired up to a standard Supabase backend. GitHub sync is available on every plan, including Free. You can connect a repo, push your code out, and run it anywhere Node runs. Vercel. Netlify. Cloudflare Pages. Railway. Fly. A five-dollar VPS in a Frankfurt datacentre. Your laptop, if you must.

The community best practice in 2026, repeated often enough that even Lovable's own staff have stopped pushing back on it, goes like this: build the first seventy or eighty per cent of a project on Lovable, where the cost-per-feature is unbeatable. When the credit burn from debugging loops, custom logic and 'why won't this one Tailwind class behave' starts to outstrip what you would spend in time doing it yourself, export to GitHub, open the project in Cursor or VS Code, and finish it like a normal piece of software. This is not a failure mode. It is the documented happy path for any project that grows beyond a prototype.

The trade-offs of leaving, honestly listed: you lose the AI Gateway for your deployed app, so any in-app AI features need their own API keys and a few lines of integration. You lose the one-click 'edit on Lovable, see it live' loop, although you can still come back through GitHub. You take on a small amount of DevOps — environment variables, build commands, occasional certificate renewals — that the platform was doing for free. And if you used Lovable's managed Supabase, you have a one-off data export and import to do when you move to your own Supabase project. None of these are difficult. All of them are work you did not have to do before.

The credit unpredictability problem nobody has solved

Lovable is not unusual in having a credit system that can spike unexpectedly. Cursor managed to make exactly the same mistake in June of 2025 and had to publicly apologise and refund three weeks of unexpected charges. v0 moved off fixed message counts to token-metered credits in May of 2025 and had its own grumpy weeks. The entire category has the same structural problem: under the hood you are paying for tokens billed by OpenAI, Anthropic and Google, and tokens are a unit that scales with whatever the model decides to read and write, which is a number neither you nor the platform can predict before the run.

Lovable's specific version of the problem is the bug loop. A developer asks the agent to fix an error. The agent introduces a new error while fixing the old one. The developer asks it to fix that. Three rounds in, fifteen credits have evaporated on a problem that a competent human would have solved in five minutes with a pair of eyes and a console log. Surveys of the developer community in 2026 put the rate of this happening at somewhere between sixty-five and seventy-five per cent on complex features. It is not a flaw unique to Lovable — Cursor, Replit and Claude Code all do it — but Lovable's flat per-message billing makes it more visible. The countermeasure is procedural: when a fix attempt fails twice, stop, switch to plan mode, ask the agent to think out loud, and only re-engage build mode once the diagnosis is concrete. This advice will save you more money than anything in this article.

The verdict, such as it is

Lovable is competitively priced for what it includes. Twenty-five dollars a month, with hosting, a real database, an integrated auth system, an AI Gateway and the bonus daily credits, is genuinely a good deal compared to its peers — provided you make it through a month without a bug loop or a traffic spike. The unified credit bank is a sensible administrative simplification that has not changed the underlying economics, only the bookkeeping. The Build mode opacity is a real and unsolved problem and the platform should be embarrassed about it. The export path is clean, the code you generate is yours, and the rational long-term strategy for any product that succeeds is to host elsewhere while continuing to build on Lovable for the parts where it remains the fastest tool in the category.

Treat the credits like petrol, not like rent. Watch the gauge. Use plan mode generously. Top up only when you must. Export to GitHub on day one so the option to leave is always there. And if anyone tells you, with confidence, that this category will look the same in twelve months — they have not been paying attention to the last twelve.