In this article
- 01 Key takeaways
- 02 One balance, four kinds of consumption
- 03 What a task actually costs
- 04 The free tier, precisely
- 05 Expiry: the part that surprises people
- 06 Top-ups, and what they reveal
- 07 What could not be verified
- 08 What a metered model optimises for
- 09 How to estimate your own number
- 10 Frequently asked questions
- 11 Sources
Lovable prices by credits, not by seats, so the headline tier tells you very little about what a month costs. What decides your bill is how many credits a task consumes, which grants refresh, and when unused credits expire. All three are documented publicly, and this article works through them as they were published on August 30, 2026.
Key takeaways
- One credit balance covers building, hosting, the built-in Cloud backend and AI features in the deployed app.
- Credit cost scales with task complexity, and the documentation gives worked examples rather than a flat per-message rate.
- Expiry rules differ by credit type, and they are the most common source of surprise.
- Top-up rates are published; subscription dollar prices were not readable on the page reviewed, so verify current pricing directly.
One balance, four kinds of consumption
The documentation is explicit that a single balance is drawn down by building your app, hosting it, running the built-in Cloud backend, and any AI features your deployed app itself uses. That last one matters and is easy to miss when you budget: an app you ship that calls AI at runtime keeps consuming from the same balance that pays for building it, so usage by your own end users lands on the same meter as your development work.
What a task actually costs
Credits are not consumed per message at a flat rate. The published examples make the spread clear: a trivial change such as making a button grey is documented at 0.50 credits, while generating a landing page with images is documented at 1.70. The practical consequence is that estimating from a message count is unreliable. Ten small refinements and three substantial generations are not the same spend, even though both are thirteen prompts.
This is worth thinking about before you adopt a working style. Iterating in many small nudges is cheaper per step but not necessarily cheaper overall, and a single well-specified generation can cost less than the ten prompts spent circling toward the same result.
The free tier, precisely
The free plan is documented as five build credits per day, capped at 30 per month. Separately there is a monthly grant of 20 Cloud credits and 4 AI credits. Those are three distinct allowances rather than one pool of 54, and each carries its own expiry behaviour. The daily allowance is generous enough to evaluate the product and deliberately not enough to run production work, which is a reasonable design and worth reading as such.
Expiry: the part that surprises people
Four rules are documented, and they differ by credit type. Daily build credits expire at the end of the day, at 00:00 UTC. Monthly grants do not roll over. General credits on a monthly plan expire two months from issue. On an annual plan they expire one month after the billing period.
Read those against the shape of your work rather than against a competitor. If your delivery pattern is a burst of building followed by a quiet month, expiring credits are a real cost that never appears on the price page. If you build steadily every week, they are close to irrelevant. The same allowance is either fine or expensive depending entirely on your workload, which is why comparing headline prices across credit-metered tools produces so little signal.
Top-ups, and what they reveal
Top-up pricing is published and gives the clearest per-unit figure available: $15 for 50 credits on Pro, which is $0.30 per credit, and $30 for 50 on Business, which is $0.60. The Business rate being double the Pro rate for the same quantity is a documented fact rather than an inference, and it is worth factoring in if you expect to buy beyond your allowance regularly.
Against the worked examples above, $0.30 per credit puts a documented landing-page generation at roughly fifty cents of top-up credit. That is a useful anchor, but do not turn it into a project estimate: it covers one generation, not the iterations, fixes and content work around it.
What could not be verified
The subscription dollar prices for Pro, Business and Enterprise are rendered client-side on the pricing page and were not readable in the version fetched on August 30, 2026. The tier names are documented; the amounts are not something this article can state. Rather than repeat a figure from a secondary source, the honest instruction is to verify current pricing on the vendor’s own page before you budget. Team structure is documented: workspace members are unlimited and draw from a pooled credit balance, so cost scales with usage rather than headcount.
For a worked contrast, Lovable and Base44 side by side covers a builder that does publish its subscription prices in full, and what that asymmetry does and does not tell you about either product.
What a metered model optimises for
Setting the unreadable amounts aside, the shape of the model is itself informative. Metering consumption rather than seats means the bill follows how much you build, not how many people are logged in — documented here as unlimited workspace members drawing on one pooled balance. For a team that builds in bursts, that is structurally kinder than paying per head for months when nobody ships. The same property inverts under sustained load: heavy build months cost more than light ones, and the ceiling is whatever you consume rather than a number you agreed in advance.
Neither shape is better in the abstract. They answer different questions. A flat per-seat plan buys budget certainty and charges you for idle capacity; a metered plan removes the idle charge and hands back the variance. The practical question is which of those two risks your finance process tolerates, and that is answerable before you know a single price.
How to estimate your own number
Take the last real project you shipped and count the substantial generations rather than the prompts — pages created, major sections regenerated, backend features added. Multiply by roughly the documented cost of a substantial generation, then add the small refinements at a fraction of that. Compare the result against a monthly allowance, and then check the expiry rules against your own delivery rhythm. That produces a far better estimate than any published comparison table, because it is built from how you actually work.
If you would rather compare a fixed-scope model, our own plans and what consumes credits on our side are set out on the pricing page, including which allowances refresh and which never expire.
Frequently asked questions
How much does Lovable cost per month?
The subscription amounts were not readable on the pricing page fetched on August 30, 2026, because they are rendered client-side. The documented tier names are Free, Pro, Business and Enterprise. Verify current pricing on the vendor’s own page rather than relying on any third-party figure, including this one.
What does one credit buy?
It depends on task complexity. The documentation gives 0.50 credits for a trivial change such as recolouring a button and 1.70 for generating a landing page with images. There is no flat per-message rate, so credit consumption cannot be estimated from prompt count alone.
Do unused Lovable credits roll over?
Not universally. Daily build credits expire at the end of the day at 00:00 UTC, monthly grants do not roll over, general credits on a monthly plan expire two months from issue, and on an annual plan one month after the billing period.
Is the free plan enough to build a real site?
The documented free allowance is five build credits per day capped at 30 per month, plus 20 Cloud and 4 AI credits monthly. That is sized for evaluation. Whether it covers a real project depends on how many substantial generations your project needs, which the worked examples above let you estimate.
Does adding teammates increase the cost?
The documentation states unlimited workspace members drawing on a pooled credit balance, so cost scales with consumption rather than with seat count. Confirm this against current documentation before you plan around it.
Written by Vincent Chen, who is accountable for the claims on this page.
Sources
Written by
Founder of PolyDraft and CEO of BlackMonolith, Inc.