Frequently asked questions
The essentials on how Semantara works, how it saves you money, and how it protects your keys and your data.
What is Semantara and what problem does it solve?
Semantara is decision infrastructure for AI: a layer that sits between your application and your providers (today OpenAI and Anthropic) and decides how each request is handled. It reuses answers you already paid for through semantic caching, enables complexity-based routing when you turn it on, and shows your spend and savings in USD. You pay less for the same answers; in your application you only change the base URL and the key.
What does BYOK ("bring your own key") mean?
You use your own provider keys. Semantara never resells tokens or marks up your usage: you keep your billing relationship with OpenAI/Anthropic, and we only charge for the platform subscription.
How does semantic caching save me money?
In workloads with FAQs, support flows, internal assistants or repetitive tasks, a meaningful share of requests often has similar intent. In those cases, every response served from cache avoids a new provider call and turns repetition into measurable savings.
How is this different from OpenAI's or Anthropic's caching?
Providers cache identical prefixes (the exact same text). Semantara caches by meaning: "how do I reset my password?" and "I forgot my login, what now?" share one answer. And Semantara stacks on top of provider caching — it doesn't replace it.
What is complexity-based routing?
When enabled, complexity-based routing lets Semantara evaluate each request and decide whether it can be handled by a cheaper model or needs a more capable one. The goal is to pay for premium capacity only when it actually adds value. It is a general platform capability, not a plan add-on.
Do I have to change my code to integrate it?
No. Semantara exposes an OpenAI-compatible endpoint (/v1/chat/completions). Change the base URL (and optionally the model) and you're done; the rest of your integration stays the same.
Which providers are supported?
Today Semantara supports OpenAI and Anthropic. Gemini is on the roadmap. Because the API is OpenAI-compatible, your application can keep a single integration while Semantara adds provider support over time.
Are my keys and data secure?
Yes. Provider keys are encrypted with AES-256-GCM, every client is isolated in a multi-tenant environment, and we never use your data to train models. See the Security page for details.
How much can I really save?
It depends on how many requests repeat, your model mix, and how reusable the answers are. Use the Savings Calculator to estimate your case; the Semantara dashboard shows savings calculated from your own requests.
Can I control my spend?
Yes. You track your usage and savings in real time from the dashboard —cost is computed on every request, per client and per key— around our core metric: dollars saved per month. And you set per-key rate limits to curb unexpected usage.
How do plans and billing work?
There's a Starter, Pro, and Business plan with fixed pricing, plus custom Enterprise. Because it's BYOK, the subscription covers the platform; model usage is billed directly by your providers. See the Plans page.
Where is it hosted? Is there an on-premise option?
The managed version runs on cloud infrastructure. For enterprise needs with sensitive data, contact us: the Enterprise plan supports custom deployments.