Zaino AI
One API for every model. OpenAI- and Anthropic-compatible endpoints, spend ceilings, and instant keys.
Trusted by people shipping with models from
One API. Every model connected.

Other tools cap your usage and throttle you mid-flow. You hit limits, wait hours, and lose momentum.
Zaino gives you a shared token pool and no per-minute throttling — so a heavy hour costs you an hour, not a day. Just code.
Choose Your Plan
Every model, one API. Pay in dinars by D17 transfer anywhere in Tunisia, or by card on Starter and Pro.
Solo
Cheapest way onto the strong models
- 35M shared tokens/month
- Spendable on Fast and Smart
- 2 API keys
- 60 requests/min
Starter
For side projects and everyday agent work
- 50M shared tokens/month
- Spendable on Fast and Smart
- 3 API keys
- 60 requests/min
Pro
For work that needs the strongest reasoning
- 50M shared tokens/month
- Spendable on Smart, Frontier and Fable
- 15 API keys
- Priority support
Team
Three times the pool, on one invoice
- 150M shared tokens/month
- Spendable on Smart, Frontier and Fable
- 30 API keys
- Priority support
Business
Company-wide rollout with finance in the loop
- 400M shared tokens/month
- Spendable on Smart, Frontier and Fable
- 100 API keys
- SSO and audit API
- Priority support
Pay by D17 transfer (Tunisia), or by card on Starter and Pro. Cancel anytime.
Frequently Asked Questions
Everything you need to know about Zaino AI and how we work.
Zaino AI is an LLM gateway for developers and teams in Tunisia and beyond. It gives you one API key and one shared token pool per plan, speaking the OpenAI and Anthropic wire formats natively so your existing tools work unchanged. Pricing is predictable, in local currency, with D17 payment support.
Each plan includes one pool of blended tokens per month, spendable on any model the plan covers — there is no per-model cap. A blended token is not a raw token: a cached re-read costs a tenth of one, an output token costs five, and the tier you use scales it (smart is the unit, frontier five times it). So a long conversation costs far less than its raw prompt size suggests, and a frontier call costs more than a smart one. Only successful inference calls are metered; failures never consume quota. Per-minute request limits apply and are listed on each plan.
Your allowance is one shared pool with no per-model limit: a heavy month on one model still leaves the rest spendable on another. When the pool is spent, every model on the plan returns a clear "out of tokens for this period" error until the plan renews — or you upgrade for more allowance immediately.
Zaino is an independent gateway built on Anthropic- and OpenAI-compatible wire formats. Compatibility means your existing SDKs and tools work unchanged against us — it does not imply any partnership with, endorsement by, or affiliation with those vendors, and we do not claim one. What we are responsible for is the layer you actually touch: the key, the quota, the meter and the billing.
Availability depends on your plan. Starter covers fast and smart. Pro adds frontier and fable, plus more API keys. Every model you are entitled to is served at full capability — no throttled variant, no smaller context window.
Absolutely. You can upgrade, downgrade, or cancel your subscription at any time from your dashboard. Upgrades take effect immediately with prorated billing. Downgrades and cancellations apply at the end of your current billing cycle. There are no lock-in contracts or hidden fees.
Every request is TLS-encrypted end to end, and we do not log prompt or response bodies: LiteLLM is configured with store_prompts_in_spend_logs off, so the spend ledger records token counts and cost, never content. We do not train on your data. The full policy is at /privacy.
Yes, when those providers are enabled on this deployment. Signing in with a provider you already use skips the password entirely, and if the address it returns has been verified by that provider it is matched to your existing account rather than creating a second one. Nothing is linked on an unverified address, and a provider email that differs from your account email never replaces it. You can keep using a password at any time — the two methods work side by side and neither is ever required.
Because not every provider verifies the address it hands us. GitHub, for example, does not verify addresses it does not own, and Apple can return a private relay address. When the address was not verified by the provider, we create the account but require our own confirmation link before it can create keys, start a checkout, use chat, or redeem a trial. That is deliberate: those actions either cost real money or spend real tokens, and we will not spend them on an address nobody has proven they can receive mail at.
Still have questions? Chat on WhatsApp or call +216 56 794 578
Ready to stop counting tokens?
One key, every model, no throttling. Start with a free trial.