Your company's AI with one key and the spend under control
Laura runs Contadores Andinos, an accounting firm. Her team wanted AI, but nobody was going to juggle keys from five providers or explain a surprise invoice. With the FinOps Gateway she got one key, loaded a prepaid balance and ten minutes later JMAX was answering with it.
Before and after
What changed at Contadores Andinos
The AI was already there. What was missing was knowing who used it, how much it cost and how to switch it off.
Before
Each person paid for their own tool. Keys from several providers, invoices in different currencies and, at the end of the month, nobody knew which project had spent what. If a key leaked, finding where it was used took a whole afternoon.
Now
One key per team, a prepaid balance and a statement with every request, its model and its cost. Each key has its own per-minute limit and its allowed models. If one leaks, Laura revokes it in one click and creates another.
Same AI. Less chaos, and a bill nobody has to explain.
How it works
From sign-up to the first answer, in four steps
Open your account
In the client portal: name, company and email. You get your own wallet.
Load your balance
Prepaid from USD 5. Today by bank transfer; PayPal coming soon.
Create your API key
One per team or project, with its per-minute limit and its allowed levels. It is shown only once.
Paste it in JMAX or your app
In JMAX: Providers → add OpenAI-compatible provider. In your code: change the base_url and the model.
Prices
Three levels of intelligence. You only pay for the tokens you use
Prices in US dollars per million tokens. Every request deducts from your balance what it consumed: input, output and cache.
USD 0.20
per million input tokens
Output: USD 0.80 · Cache: USD 0.05
Classification, data extraction, translations and short summaries, at high volume.
- Text
- Lowest cost per token
- Ideal for repetitive tasks
USD 1.50
per million input tokens
Output: USD 6.00 · Cache: USD 0.30
Explanations, writing, code and analysis: the balance between quality and cost for everyday work.
- Text and tools (web search, documents)
- What JMAX uses day to day
- Adjusts power to each request
USD 4.00
per million input tokens
Output: USD 16.00 · Cache: USD 0.80
Deep analysis, architecture, mathematics and multi-step reasoning.
- Text, tools and reasoning
- For the hardest questions
- Same key, just change the model
Prices published by the gateway on 7 October 2026. If a provider fails, the request continues with an equivalent engine at the same price.
Calculator
How much would your team spend a month?
Move the bar: queries per working day across the whole team.
100 queries a day
Estimate for a typical query of 3,000 input tokens and 600 output tokens. A query with web search reads several pages and can cost five times more; your portal shows the real figure for every request.
Who uses it and when
Built for a team that already wants to use AI
The partner or manager buys it
Opens the account, loads the balance and creates one key per team. Decides which levels each key may use.
The team uses it in JMAX
Without seeing any key: JMAX keeps it, and personal data travels replaced by markers.
Developers, in their own app
Any OpenAI SDK works: Python, JavaScript or plain HTTP. Change the base_url and the model.
When starting with AI
Load USD 5, try it for a week and look at the statement before scaling up.
When giving a team a budget
A separate key with its own limit: marketing does not eat the accounting budget.
When a key leaks
Revoke it or rotate it from the portal; the others keep working.
Questions
What people ask us
How is it billed?
It is prepaid. You load a balance and every request deducts the cost of the input, output and cache tokens it consumed, at the published price of each level. No monthly fees or contracts, and every movement is in your portal.
What happens when my balance runs out?
New requests are rejected with a clear error until you top up. There are no debts and no surprise charges.
Do I need JMAX?
No. The gateway speaks the OpenAI format, so it works with any application. With JMAX it is just quicker: you paste the key and assign it to the roles.
What about the privacy of my data?
Requests are processed only to produce the answer. We record usage metrics (tokens, cost, latency) for your statement. Keys are stored encrypted and shown only once. From JMAX, personal data travels replaced by markers.
Why "request access" and not "create account"?
We are opening the client portal by invitation. Leave us your details and we activate your account with your first balance; we answer within one working day.
Your team, with its own key this week
Tell us how many people would use it and for what. We activate your account in the client portal and help you set it up in JMAX.