Lovable

Where Lovable sits

Lovable is a chat-first app builder: you describe a web application in plain language and it produces a working front end, a database, authentication and a deployed URL without you writing the scaffolding by hand. Founded in Stockholm in 2023 by Anton Osika and Fabian Hedin, the company has become the reference point for what the industry calls vibe coding — a label I use reluctantly, because it undersells what the product actually does. At time of writing Lovable is a private company that raised a $400 million Series C at a $13.3 billion valuation in August 2026, with reported annual recurring revenue approaching $600 million (Business Insider; TechStartups, August 2026). Those numbers matter for a buyer: this is a well-funded, fast-moving product with a large user base, not a weekend project that might vanish.

The pitch is that a non-engineer can ship a real, database-backed web app — and that an engineer can use the same surface to move faster. In my testing both claims hold, with caveats I set out below. The honest summary is that Lovable is very good at generating the 80% of a CRUD application that is tedious to write, and considerably less reliable on the 20% that is architecturally demanding.

What Lovable actually does

The workspace is a project chat with three modes. Plan mode discusses an idea without touching code and costs 1 credit per message plus any subagent research it runs. Build mode implements changes and charges by the complexity of the request — Lovable's own pricing page gives worked examples at time of writing: changing a button style costs 0.50 credits, removing a footer 0.90, adding authentication 1.20, and a landing page with images 1.70. Chat mode is a cheaper, separate per-message model for questions that do not modify the project (docs.lovable.dev, credits and usage).

Underneath, Lovable generates React front ends, wires a Supabase project as the built-in backend (Postgres, auth, storage, realtime), can sync the code to GitHub so you own the repository, and hosts the deployed app on its own Cloud runtime. One credit balance pays for building, chatting, hosting and the AI gateway calls your deployed app makes to models. That single-balance design is both the product's elegance and its main consumer trap: a busy app burning AI gateway credits draws from the same pool as your build work.

Pricing at time of writing (lovable.dev/pricing and docs.lovable.dev/introduction/subscription-plans): Free at $0 with 5 build credits a day capped at 30 a month, 20 Cloud credits and 4 AI credits a month. Pro starts at $25/month for 100 monthly credits and scales in tiers to $2,250 for 10,000. Business starts at $50/month for the same 100 credits — roughly double Pro at every rung — and adds SSO, role-based access, internal publishing and data excluded from model training by default. Enterprise is quoted. Pro top-ups cost $15 per 50 credits; Business top-ups cost $30 per 50. Students and teachers get 50% off, but only on the Pro plan with 100 credits on monthly billing.

My testing notes

Two things stood out in my testing. First, the speed from prompt to deployed prototype is genuinely fast. A single Build mode request for a landing page with a hero, pricing section and contact form came back as a working, deployed page in a few minutes, with the Supabase tables created alongside it. The edit loop — selecting something wrong on the page and describing the fix — is the strongest part of the product. When I asked for a style change to a button, the 0.50-credit charge matched the pricing page's example almost exactly, and the change landed without collateral damage to the rest of the layout. Iteration feels closer to commenting on a mockup than to writing code.

Second, credit accounting is where trust erodes. On the Free plan I hit the monthly ceiling early: 5 build credits a day reach the 30-credit monthly cap on the sixth day, after which the daily grant stops until the next calendar month. The docs are explicit about this and about the fact that daily build credits expire at midnight UTC and do not roll over. Even so, watching a single multi-step request consume several credits — exploration, verification and browser checks all count — makes budgeting approximate. I learned to use Plan mode for anything with architectural consequences: at 1 credit it is cheap insurance against spending 10 credits rebuilding a data model the tool got wrong.

Third observation, on code ownership: the GitHub sync is the feature that separates Lovable from toy generators. In my testing the exported repository was a conventional React project I could clone and run locally, which means the escape hatch is real. That said, the Supabase wiring assumes you stay in the Lovable-shaped architecture; pulling the project into a different backend is manual work.

Where it falls short

The limits are consistent across my tests. Complex, multi-step features with strict invariants — payment state machines, role-based permissions with edge cases, background job orchestration — need several rounds of correction, and each round costs credits. The tool is confident in its output; it will produce something that looks finished while silently skipping a requirement. I caught it omitting validation logic I had explicitly asked for, and the fix required pointing at the exact handler.

Credit economics also cut against heavy users. Business at double Pro's price buys governance features, not more building; if you are a solo builder spending heavily, the effective hourly cost of iteration can approach that of a contractor for small jobs. Free plan users cannot top up at all. And there are fewer guardrails than an engineering team would want: no formal review gates, no test generation you can rely on, and AI-generated code that can drift in style between iterations.

Vendor claims about generated app quality should be read as vendor claims. Lovable says daily Cloud and AI grants are temporary and subject to change — a rare bit of pricing candor, but it means the free economics I tested may not last.

One further observation is about collaboration and deployment behaviour. Lovable lets unlimited collaborators into a workspace even on Free, which I did not expect at that tier, and the real-time co-editing model means a founder and a designer can sit in the same project chat. In my testing the deployment step itself was uneventful: the app went live on a lovable.app address on Free, and custom domains are a paid-plan feature. What I could not verify is how the Cloud runtime behaves under real production load — the docs describe database, network, storage, compute and realtime usage as billable Cloud consumption, but there is no published uptime commitment on the self-serve plans. Teams shipping anything with contractual availability requirements should ask about that before they rely on Lovable hosting as their production surface.

Verdict

Lovable is the most complete chat-to-app product I have tested for the classic founder use case: validate an idea, get a real database behind it, put it in front of users this week. It is rated 4.1 on the ToolSpotter rubric — depth of core functionality carries most of that weight, because full-stack generation with Supabase, GitHub export and hosting in one loop is genuinely deep; value and usability are strong at the $25 entry price but dented by credit burn; the ecosystem around it is large and the company's funding and ARR are strong reliability signals. If your project has hard correctness requirements, treat Lovable as a fast prototyping layer with a manual hardening pass, not as an autonomous engineering team.

Sources

• https://lovable.dev/pricing
• https://docs.lovable.dev/introduction/credits-and-usage
• https://docs.lovable.dev/introduction/subscription-plans
• https://en.wikipedia.org/wiki/Lovable_(company)
• https://businessinsider.com/lovable-funding-valuation-doubles-13-3-billion-vibe-coding-2026-8
• https://techstartups.com/2026/08/12/swedish-vibe-coding-startup-lovable-raises-400-million-at-13-3-billion-valuation-as-arr-nears-600-million
• https://empyre.dev/articles/lovable-pricing
• https://builddu.com/guides/lovable-pricing
• https://agentsite.ai/blog/lovable-pricing
• https://revex.agency/blog-posts/lovable-pricing-guide-2026