Hold → settle → refund
Each request reserves a conservative hold. When the stream ends we charge the exact tokens the provider reports and release the remainder instantly. If the provider fails, the whole hold is released.
Token gateway
One key.
One key.
Every model.
Zero surprises.
ZettaForge sits between your app and the models you choose. We meter every token from the provider’s own usage record, hold credit before each call and publish every fee line-by-line. No silent markup, no surprise balance drains.
OpenAI-compatible
3% credit fee, shown before you pay
Free tier on signup
—models live today
0%markup on tokens
3%credit fee, itemised
1endpoint to integrate
Why teams switch
Metering you can audit,
Metering you can audit,
limits you can trust.
Most resellers estimate spend and reconcile later. We reserve the worst case before the call, settle on the provider’s reported usage and refund the difference in the same second.
Per-key budgets
Set daily and monthly caps per API key. Keys pause themselves the moment a cap is reached.
Security by construction
Argon2id passwords, rotating refresh-token families with reuse detection, hashed API keys, HMAC-signed webhooks with a 5-minute replay window, and a strict Content-Security-Policy. Upstream hosts come from the operator’s catalogue, never from the request.
Usage you can prove
Every call is logged with model, tokens, cost, latency and whether the usage figure came from the provider or an estimate. Export 30 days of CSV whenever finance asks.
Catalogue
Prices pulled from the
Prices pulled from the
live catalogue.
Prices below are per million tokens at the current sale rate, read straight from the live catalogue.
Model
Input / Mtok
Output / Mtok
Context
Integration
Three steps. One base URL.
Point any OpenAI SDK at https://your-domain/v1. Your model names, tools and streaming calls work unchanged.
Create a key
Sign up, open API keys and generate a key. The secret is shown once and stored only as a hash.
Set a budget
Give the key a monthly cap so a runaway loop cannot drain your balance.
Swap the base URL
Replace api.openai.com with our host. Done.
Pricing
Pay for credits.
Pay for credits.
Not for seats.
Token prices are the provider’s list price plus our published markup. Credits are bought in one go and carry a single fee, shown in full before checkout.
3%platform fee on top-ups
- Unlimited API keys, playground and arena
- Exact per-token metering with refunds
- Daily and monthly caps per key
- CSV export and audit trail
- No monthly minimum, no lock-in
Top-up calculator
See exactly what lands in your balance.
Payment$50.00
Platform fee (3%)$1.46
Credits added$48.54
Questions
Straight answers.
Anything else? Write to hello@zettaforge.dev.
Is this the same price as calling providers directly?
The per-token price is the provider’s published list price plus the catalogue markup shown on this page. Our revenue comes from the single fee on credit top-ups, itemised at checkout. Nothing is added to individual requests.
How do you meter usage?
We request the provider’s usage block on every call, including streams. If a provider omits it, we estimate conservatively and label the entry estimated in your usage log.
What happens if a model errors mid-stream?
You are charged only for tokens already delivered. The remainder of the hold is released immediately, and the entry is marked in your log.
Do you train on my data?
No. Prompts are proxied to the provider you select and are not stored beyond the usage counters. Provider retention depends on that provider’s policy.
Can I limit spending per key?
Yes. Each key accepts a daily and monthly cap in USD. When either cap is reached the key pauses itself and returns 402 until the next window.
Ready when you are
Prototype today,
Prototype today,
ship on one key.
Create an account, generate a key and send your first request in under two minutes.