Bolt.new Pricing: How Many Tokens a Business Tool Really Uses

Bolt.new has a Free plan with 1 million tokens a month and a 300,000-token daily cap. Pro starts at $25 monthly with 10 million tokens and no daily cap; Teams starts at $30 per member monthly. A simple business tool has no fixed token price because later prompts can read its growing project. Measure each build and correction.

Bolt's own QuickStart illustrates the spread: its sample first build used about 204,000 tokens, while a later title change used about 70,000. Those figures belong to that example, not to your tool. The useful budgeting unit is a sequence of prompts through a tested release, including the corrections owners often forget to count. Hosting is listed on the plans, but ongoing site use has its own limits.

Follow me on Instagram@sagnikteaches

The current plans and the token meter

Bolt's public pricing page shows Free at $0, Pro at $25 per month billed monthly and Teams at $30 per member per month billed monthly. Free includes 1 million tokens monthly but no more than 300,000 in a day. Pro starts with 10 million monthly tokens and removes the daily limit. Teams adds central billing and team controls; its subscription is per member. Bolt also offers yearly billing, which its page says saves up to 28%, so use the displayed monthly rates only for a monthly-billing comparison and verify the yearly checkout separately.

Connect on LinkedInSagnik Bhattacharya

“Token” is a unit of model processing, not one word you type. Bolt's pricing FAQ says it spends tokens when it reads your prompts and project files, makes a plan or writes code. A short instruction on a larger project can therefore spend more tokens than a long instruction on an empty project. The project grows as pages, data handling and libraries are added. A price based on counting characters in your prompt alone will mislead you.

Subscribe on YouTube@sagnikteaches

Free includes website hosting and a bolt.host address, according to Bolt's introduction. Paid plans offer custom domains and more publishing options. The pricing page also lists web-request allowances: Free up to 333,000 and Pro up to 1 million. These are not the same meter as build tokens. A tool might have plenty of AI build tokens left yet need a review of hosting usage, database capacity or a separate external service. Price those dependencies before exposing a tool to staff or customers.

Define a tool small enough to cost honestly

Use one job for the first build. An illustrative signage company wants a quote calculator for standard banners. A staff member enters width, height, material, finishing and quantity. The tool calculates an indicative price from an approved rate table, displays the components and marks non-standard work for a manual quote. It does not need customer login, payment collection, live inventory or automatic email on day one. Each added feature creates code and test paths, which can increase token use on every later prompt.

Write down the maths before opening Bolt. Suppose the illustrative rate card charges $18 per square metre of material, $6 finishing per item, a $12 setup charge and a minimum selling price of $35. For a 2 × 1 metre banner with finishing and quantity two, material is 2 × 1 × $18 × 2 = $72, finishing is $12, setup is $12 and the indicative total is $96. For a 0.5 × 0.5 metre single banner without finishing, material is $4.50 plus $12 setup, so the $35 minimum applies. These are sample business rules, not vendor facts.

Give Bolt an acceptance table rather than “make a calculator”. Include at least a normal case, a minimum-price case, a zero or negative dimension, an unsupported material and a quantity threshold requiring a manual quote. The values tell you whether the generated logic is right. Without them, a polished screen can conceal an incorrect formula. The Bolt quote-calculator tutorial focuses on the build itself; the question here is how to measure and control the token cost of that build.

A token budget for the banner calculator

The following run is explicitly illustrative. It is a planning envelope, not an observed Bolt charge. Imagine the owner checks the subscription balance before and after each phase. The first prompt creates the input form and calculation; two follow-ups correct the minimum-price rule and invalid-input handling; a final prompt improves the result summary and publishes the tool. The recorded numbers below are invented to show the arithmetic you should perform with your own balance.

The monthly total fits Free's 1 million tokens, but the 220,000-token first build leaves only 80,000 under the 300,000 daily cap on that day. An 85,000-token fix would need to wait for the daily allowance to refresh or use a paid plan. This is what “runs out” first in a Free build: the daily cap can interrupt a project that still has ample monthly tokens. If the first build had used 290,000, even a small correction might be postponed.

Bolt's actual QuickStart shows about 204,000 for its first app build and about 70,000 for a simple title edit in its own example. That makes the illustrative 220,000 and 85,000 phases plausible planning numbers, but does not validate the whole 580,000 estimate. A calculator with different dependencies or a different model could behave differently. Always record the subscription balance and token display in chat, as Bolt's QuickStart instructs, instead of quoting this example as a forecast.

If a five-person office pays for Teams, do not multiply the members' allowances into one shared construction budget. Bolt says each paid member gets their own allotment and tokens are not shared. Perhaps one staff member builds the calculator while four only review it. The seat cost at the displayed monthly rate is five × $30 = $150 a month, but the builder's remaining tokens are still the relevant constraint. Bolt also lets a team add free members who can view team projects but cannot send prompts, because they get no tokens, so the four reviewers may not need paid seats at all. Buy Teams for collaboration and administration, not because five small balances seem to form one large balance.

Put an acceptance brief beside the token log

A build can consume a modest number of tokens and still be unusable. Put the rules in a one-page acceptance brief before the first prompt. For the illustrative signage calculator, define the input units, allowed materials, finishing charge, setup charge, minimum price and the cases that need a manual quote. State whether the result is internal or visible to customers. An owner can then compare each Bolt output against a stable specification rather than altering the goal after every attractive demo.

Here is a concise illustrative brief in ordinary language: “Use width and height in metres. Accept positive numbers only. Material costs $18 per square metre. Finishing adds $6 per item. Setup adds $12 per order. Apply a $35 minimum after calculating the total. Quantity over 20 or unknown material shows ‘Manual quote required’ and no firm price. Show the calculation to staff.” A later change should say exactly which rule is being changed and which examples must still pass.

Keep a test ledger with expected and actual output. The normal 2 × 1 metre, quantity-two case should show $96. The 0.5 × 0.5 metre, quantity-one case should show $35. Width zero should show an input error, not a $35 quote. An unknown material should show “Manual quote required”, not silently use the standard rate. Quantity 21 should take the manual route. After each AI change, rerun all five cases. A fix that repairs the minimum but breaks the standard case has not passed, even if the token meter shows efficient use.

Use that ledger to decide whether another prompt is worth its cost. If one button label is wrong but the maths is correct, a direct text edit or later maintenance pass may be enough. If the formula produces $84 instead of $96, stop and correct it before publishing. Record “85,000 tokens; minimum fixed; standard case still correct” rather than “85,000 tokens; looks better”. The log now connects usage to delivered value.

Measure token use without guessing from the prompt

Before a prompt, open Settings from your profile icon and take the remaining-token figure from Subscription & Tokens under Account. Then turn on “Display token usage in chat” under Account, General, as Bolt's QuickStart describes. After the prompt finishes, record the new balance and the displayed consumption. If the balance changed by more than expected, note the project size, model choice and what Bolt built. Keep a small log: date, prompt goal, before balance, after balance, output, accepted or rejected. This separates useful iterations from repeated attempts that did not improve the tool.

Test the product between prompts. Enter the 2 × 1 metre case and confirm $96. Enter the small case and confirm $35. Enter width zero and confirm there is no quote. If a test fails, ask for one specific correction with the expected output. “Fix the pricing” invites Bolt to revisit a broad set of files and may introduce a new error. A bounded instruction such as “For the 0.5 × 0.5 metre case, apply the $35 minimum after setup and finishing; keep the 2 × 1 metre case at $96” gives a clear target.

An illustrative mistake is asking Bolt to “add tax, payments, inventory and customer accounts” in one follow-up. The project may grow dramatically, while the owner can no longer tell which change caused the price error. Split business rules from infrastructure features. Confirm whether staff need an internal estimate, a customer-facing quote or a paid order before adding payment processing. A different tool may be more suitable if the intended workflow already belongs in a CRM or quoting system.

Bolt's QuickStart also shows that restoring a version did not spend tokens in its example, and that a simple code-view text change can avoid an AI prompt. Those options are useful only if you understand their effect and verify the result. A non-coder should not hand-edit pricing logic merely to save tokens; a silent arithmetic error can be much more expensive than one measured correction. For wording, spacing or labels, direct editing may be reasonable after saving a backup.

Plan for fixes after the first publish

A live calculator needs an owner for rate updates, exceptions and faulty output. Reserve token capacity for the first month of real use. Suppose the illustrative build used 580,000 tokens on Free and the office has 420,000 left that month. If two unexpected fixes each consume 160,000, the balance falls to 100,000. That may be enough for small wording changes but not for another substantial feature. The owner should decide whether to wait for the next allowance, buy Pro or reduce scope.

Pro starts at 10 million monthly tokens and removes Free's daily cap. Bolt says unused tokens on a paid plan stay in the balance for up to two months after they arrive, provided the subscription stays active. This can help a project that alternates between heavy build weeks and quieter maintenance, but it is not indefinite banking. A tool that needs only one burst of work may not justify paying for Teams month after month. The non-coder Bolt assessment is useful for thinking through ownership after the novelty of the first build.

Separate build cost from running cost. The pricing page includes hosting, but lists web requests and database capacity as plan allowances. If the calculator is internal and used by ten staff a few times a day, the hosting picture differs from a public calculator promoted to thousands of visitors. Check the current plan's request and storage meters, and identify any external service that bills separately. The token total in a chat is not an all-in operating price.

Before public launch, check security and business accuracy. The sample calculator should not accept customer secrets it does not need. A quote should say whether it is indicative and when a person will confirm it. Test that a visitor cannot change the rate table through ordinary page inputs. Verify the live published version, not only the preview. The generated-app security checklist covers the wider release review that token arithmetic alone cannot provide.

Five ways a small build can consume more than planned

First, the initial request may quietly include account creation and a database. Bolt's QuickStart includes sign-up, login and favourites in its sample app and notes that those features make Bolt set up databases, which takes longer. A read-only calculator needs no account. If the owner later needs saved quotes, specify who can see them and why before accepting the added complexity. Second, repeated vague requests invite broad rewrites. Use observed failing inputs and expected outputs to narrow each repair.

Third, large imported files can enlarge the context Bolt reads. Bolt's pricing FAQ says it spends tokens reading project files as well as prompts, and its QuickStart says bigger, more complex apps take more tokens to build. Remove irrelevant sample assets and avoid pasting long policy documents into a prompt when a five-line rule will do. Fourth, polishing can become endless: button colour, animation, icon shape and marketing copy may each trigger a new AI pass. Agree a release threshold before starting. Functional accuracy and a usable mobile layout matter more than a tenth colour revision.

Fifth, a team can mistake published hosting for a finished operating process. In an illustrative solar installer, a roof-area estimator returns a plausible number but nobody updates material rates when suppliers change. The tool becomes wrong while still available at its URL. Name a rate-table owner and review date. A simple spreadsheet might be easier to maintain if the team cannot support the generated tool. The Bolt and Replit comparison can help if maintainability becomes the deciding factor.

The buying choice follows the measured run. Try Free for a bounded proof of concept and expect its daily cap to interrupt some build days. Move to Pro if the real project needs more iterations or custom-domain publishing. Choose Teams when several builders need shared administration and separate paid seats. Before each upgrade, take a fresh subscription-balance screenshot, list the remaining acceptance tests and calculate how many iterations you are actually buying room for.

Three short scenarios that change the plan choice

An interior designer wants a private materials selector for two staff. It reads a fixed list of finishes and returns a room estimate; no client records are stored. The owner can build and test it on Free, waiting for the next day if the daily cap interrupts corrections. Pro becomes relevant if the build needs a concentrated burst of revisions. Teams adds little unless both staff need to edit the project under shared administration.

A security installer wants a customer-facing appointment request tool. The first version works, but it also collects access details that should never appear in a public form. The next step is not buying tokens for more features. Remove the sensitive field, test data handling and confirm the published version before accepting enquiries. If the company needs authenticated customer records and service history, scope that as a separate system decision with proper access review.

A solar installer already has an estimator in use and only needs to change its rate table monthly. An AI prompt to regenerate the whole app each month is a poor maintenance plan, regardless of whether 10 million tokens are included. Design a simple, controlled rate update and test the old and new calculations. The relevant cost is the time and risk of maintaining the tool, not the unused token balance shown in an account.

Bolt token questions after your first build

Does publishing a Bolt project need a paid plan?

Bolt says Free and paid users can publish to a bolt.host subdomain, and the pricing page lists website hosting on Free. A custom domain is a paid-plan feature. Hosting still has plan allowances such as web requests and database capacity, so check those separately from the AI token allowance for building and editing.

Do unused Pro tokens roll over indefinitely?

No. Bolt's pricing FAQ says unused tokens from a paid plan's monthly allowance stay in your balance for up to two months after you receive them, as long as the paid subscription is active; Free tokens do not roll over at all. Treat the following month as the last chance to use that balance, and check Subscription & Tokens rather than adding several months of leftovers into a permanent reserve.

Can a Teams member spend another member's tokens?

Bolt's pricing FAQ says Teams is charged per member and each paid member receives their own monthly token allotment; tokens are not pooled across members. If one person builds most of the tool while others mainly review it, model the builder's allowance separately instead of multiplying the team total and treating it as one shared pot.

Further reads

• Best Workflow Technology for Small Businesses: Add AI to What You Own — Decide whether a custom tool is the right format.
• Botpress Pricing Plans: Free, Plus and Team for a Small Business — Contrast token pricing with conversation pricing.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Bolt pricing page and pricing FAQ; Bolt support documentation: QuickStart guide and Intro to Bolt (checked 5 November 2026).