kickllm / Margin Studio

FROM A COST ESTIMATE TO A BUSINESS DECISION

Your AI has a cost.
Does it have a margin?

Turn a usage CSV into a clear cost breakdown. Test revenue, growth and cost assumptions. Take an explainable report into your next pricing decision.

Runs locally in your browser · Bring your own rates · No API keys

DECISION PREVIEWIllustrative data

30-day API cost → contribution after selected costs

$600 API cost
API $600 Other $200 Contribution $1,200

Example only. Excludes costs you have not entered. Not accounting profit.

01 / Import a normalized CSV02 / Inspect costs & assumptions03 / Export a decision report

SEE THE DECISION BEFORE YOU BUY

What changes when usage grows?

This limited preview uses invented numbers. Adjust the assumptions to see the tradeoff; it does not analyze your files.

Baseline: $600 API cost over 30 days. Constant rates and token mix. Growth scales API cost; revenue and other costs stay as entered.

LOCAL DEMO · SYNTHETIC DATA
Monthly API cost$600.00
Contribution after selected costs$1,200.00
Contribution margin60.0%
Revenue needed at target margin$2,666.67

Revenue needed = (API + other costs) ÷ (1 − target margin).

Open the sample report ↗

A FINITE JOB. A FILE YOU KEEP.

Bring the usage. Keep the explanation.

01

Find the cost concentration

Group modeled API costs by project and model. Keep entered provider charges separate from calculated token estimates.

02

Pressure-test your pricing

Change revenue, usage growth, other costs and target margin. Model hypothetical token and rate reductions with explicit assumptions.

03

Leave with a report

Export a portable decision report with the cost breakdown and caveats. Your analysis runs locally, without an API account.

Check the format first

The app accepts a normalized CSV. Use the included preparation tool to map source columns and confirm cache semantics before import. All rates are USD per million tokens, supplied by you.

Download an illustrative CSV ↓
project, model, input_tokens, output_tokens,
input_per_million, output_per_million

Optional: cached_input_tokens, cache_read_per_million, requests, actual_cost_usd. Input tokens include cached input. Cache writes, tools, image/audio pricing and tier rules are not separately modeled.

KICKLLM MARGIN STUDIO

Make the next pricing
decision with your data.

For developers and small teams with API usage to audit and a revenue decision to make.

If you only need a quick estimate, the free KickLLM calculator may be enough.

$39 USD · one time
  • Offline HTML application in a ZIP
  • Usage audit and per-project contribution plans
  • CSV column preparation tool
  • Conditional cost-reduction scenarios
  • Measured cost, acceptance and latency review
  • Portable report export
  • Sample CSV and format guidance

Checkout status is being checked. No payment has been started.

14-day refund requests · Support: support@zovo.one. Internet required for Stripe checkout and download. No subscription or live rate feed is included.

Review license, privacy & purchase terms ↗

BEFORE YOU DECIDE

Know exactly what you are getting.

How do I open it on a Mac?

Download the ZIP, double-click it to extract, and open the included index.html in a modern browser. The analysis app works locally after download. The README contains format and usage instructions.

Does it know current model prices?

No. You supply rates in the CSV. Sample rates are invented examples, not provider quotes. Verify the rate, units, cache semantics and applicable discounts against your own provider and billing period.

Will it reconcile exactly to my invoice?

Not necessarily. Entered provider costs and modeled token costs are displayed separately. Taxes, credits, negotiated discounts, tool calls, audio/images, cache writes and pricing tiers can cause differences. It is an analysis aid, not an accounting system.

Does it recommend a cheaper model with the same quality?

No. Lower token or rate assumptions create hypothetical scenarios. The included evidence companion summarizes your measured trial costs, acceptance and latency, checks task coverage, and applies your own comparison gates. You must still validate representativeness, eligibility and implementation cost before changing production behavior.

Does my usage CSV leave my computer?

The offline analysis app processes your CSV locally in your browser. Checkout and download use the payment server and Stripe; those services process purchase information. The storefront preview uses synthetic data and does not request your CSV.

Can I use any provider CSV?

You can use any model name with explicit rates, but the columns must match the documented normalized format. The included preparation tool supports manual column mapping and fixed rates; native provider schemas are not automatically verified. Check the downloadable sample before buying.

Is this a hosted monitoring service?

No. It does not connect to production APIs, enforce budgets, synchronize accounts, or monitor spending in real time. It analyzes the files and assumptions you provide.