BYOK: Why We Charge ₹299 for Unlimited AI Queries

Here is a pricing detail that confuses people the first time they see it: our Pro plan costs ₹1,199 a month and caps out at 2,000 queries. Our BYOK plan costs ₹299 a month and has no query cap at all. The cheaper plan does more. That looks backwards until you understand what BYOK actually removes.

What “bring your own key” means here

Every time your chatbot answers a question, something has to call an LLM — Groq's llama3-8b-8192 or OpenAI's gpt-4o-mini, depending on what you choose. On every other plan, that call goes through our account, and we pay for it. On BYOK, you paste in your own Groq or OpenAI API key. Your chatbot still runs on Minyut's chunking, retrieval, and widget — the LLM call at the end just goes out under your key instead of ours.

We encrypt the key with AES-256-GCM the moment you save it, and it is only decrypted in memory, inside the edge function, at the exact moment a query needs it. The plaintext key never touches a log, a database column you can read, or a frontend file. You can see the setup at the BYOK docs.

Why the queries are unlimited

On Starter and Pro, we are the ones paying the LLM provider for every query, so we cap query counts to keep our own costs predictable. On BYOK, we are not paying for a single LLM call — you are, directly, at whatever Groq or OpenAI charges you. There is nothing left for us to meter. The ₹299 a month is not a discounted version of Pro. It is the price of the platform with the token cost removed entirely, because that cost is no longer ours to charge for.

What ₹299 is actually paying for

Not the AI. The AI is now a direct relationship between you and Groq or OpenAI, billed on their terms, at their rate. The ₹299 covers everything Minyut still does around that call: chunking and embedding your documents, hosting the vector index, running the widget on your site inside an isolated Shadow DOM, serving the dashboard where you manage chatbots and read query logs, and keeping five edge functions live with no cold starts. That is the platform. The tokens were never really the product — they were always a pass-through cost we happened to be marking up on every other plan.

This is also why BYOK is priced flat instead of usage-based on our end: our cost to run your chatbot barely changes whether you send 200 queries a month or 20,000. The variable cost that used to scale with usage — the LLM bill — now scales on your key, not ours.

When BYOK is worth switching to

If you are comfortably inside the Free or Starter query limits, there is no reason to switch — BYOK adds a setup step (getting an API key from Groq or OpenAI) for no benefit yet. The moment you can see yourself consistently approaching or exceeding the 2,000-query Pro ceiling, or you simply want to stop thinking about a cap altogether, BYOK becomes the cheaper, simpler option — not despite the lower price, but because of what that price no longer includes.

It is also the right move if you would rather see your AI costs itemized on your own Groq or OpenAI invoice than bundled into a SaaS subscription. Some founders want that visibility. BYOK is built for exactly that preference.

How switching actually works

You do not rebuild anything to move to BYOK. Your chunked documents, your widget, your appearance settings, and your query history all stay exactly as they are — BYOK only changes which account pays for the LLM call at query time. In practice: create a Groq or OpenAI account if you do not already have one, generate an API key, paste it into your Minyut dashboard, and pick which model to use. From your next query onward, that call runs on your key.

Groq tends to be the faster and lower-cost of the two to start with if you have not used either provider before. OpenAI's gpt-4o-mini is the option if you specifically want GPT-family responses. Either way, keep an eye on your provider's balance the same way you would watch any prepaid account — that part is now yours to manage, not ours.

The honest tradeoff

BYOK asks you to hold an API key and keep a small balance funded with your provider. If that balance runs out, queries fail until you top it up — that responsibility sits with you, not us. In exchange, you get a cost structure that scales at wholesale rates instead of subscription rates, and a query count that never runs out.

We built it this way because it is the honest version of the pricing story: charge for the platform, let the model cost go where it actually belongs. If that tradeoff fits how you want to run your chatbot, compare all four plans side by side, or start free and switch to BYOK later — nothing about your chatbot, documents, or widget changes when you do.