Rana API
One authenticated entry point into every Rana AI capability.
Every Rana product — Rana Brain, Rana Dive, and every Rana Plug-in — is reachable through the same API. One key, one integration point, and the optimization, grading, and cost-control layer underneath every one of them comes along for free.
Access
Authentication & access control
API key authentication
Every request authenticates with a Bearer API key issued to your account. Simple, standard, and how every Rana integration in production works today.
Customer sub-keys
Issue scoped sub-keys under your own parent key so you can offer Rana-powered features to your own customers — each sub-key's usage is tracked and capped independently of the others, and independently of yours.
OAuth 2.0 (planned)
OAuth 2.0 client-credentials support is on the roadmap for teams that need it as part of a broader identity setup. API-key auth is what's live today.
Usage & billing
Track it, cap it, bill it — down to the customer
Prompt request logging
Full request and response capture is available per key, encrypted at rest, so you have a real audit trail when you need to see exactly what a call sent and got back.
Per-customer usage tracking
Every call is attributed down to the sub-key and end user that made it, so "how much did this one customer actually use" is always a real number, not an estimate.
Bill your own customers
Usage is broken out per sub-key, so you can invoice your own customers for exactly what they consumed instead of eating the cost or guessing at a flat fee.
Set your own markup
Add your own margin on top of usage costs when you resell metered AI access — the economics are yours to set, not fixed by us.
Budget caps that actually hold
Hierarchical spend and request caps — company, sub-company, and individual user — are enforced before a call goes out, not after the invoice arrives.
Under every call
The optimization layer
Prompt Engineering Optimization
Vague or underspecified prompts get rewritten against four building blocks — Role/Context, Task, Audience, Format — before they ever reach a model. Optional, and off by default when exact wording matters.
Model & Prompt Route Optimization
Every request is routed to whichever model in a real, evaluated pool actually fits that specific kind of task, instead of a single fixed model handling every call.
Model Cost Optimization
Routing weighs quality against price, so you're not paying frontier-model rates for a question a smaller, cheaper model could have answered just as well.
Response grading
Every response is graded before it's billed. One that grades below quality threshold is a $0 chargeback, and it's automatically retried on a stronger model.
Provider fallback
An automatic fallback chain moves to the next-best model when a provider errors or rate-limits, so an outage on their end doesn't become an outage on yours.
Built-in abuse & safety moderation
Every request passes a moderation gate before it ever reaches a model — the same gate, whether the caller is a paying customer or a public demo visitor.