Home · Pricing

Pricing

Effective September 24, 2026

Downloading and installing the app is free. Every plan includes the agent with all its tools and the plan models: Gemma 4 31B, GPT-OSS 120B, GPT-OSS 20B, Nemotron 3 Nano 30B, Nemotron 3 Super, Nemotron 3 Ultra. Requests to them cost nothing within your plan's monthly token allowance; the allowance is shared by all plan models and resets on the 1st of every month, Moscow time. The plan models are provided by an external provider, Ollama (the Ollama Cloud service, ollama.com), and the text of requests to them is sent to it. Paid plans also give access to other models (Claude, GPT and more), charged to your ruble balance at purchase cost plus a small margin that covers payment processing fees and taxes.

You subscribe in the app for a month or a year, and the subscription renews automatically until you turn renewal off; the paid period stays active when you do. Once the monthly token allowance is used up, the plan models are unavailable until the 1st of the next month, while the other models of a paid plan keep working from your balance.

How request cost is calculated

Models process text in small chunks called tokens (one token is roughly 4 characters of English text or 3 to 4 characters of Russian). The cost of a request to a paid model is the number of input tokens (your task, the files the model reads, the conversation so far) multiplied by the input price, plus the number of output tokens (the model's reply) multiplied by the output price. Prices below are in Russian rubles per 1 million tokens and are not subject to VAT. Requests to the plan models are sent to their provider, Ollama (ollama.com), and requests to all other models are sent to their developers; both are processed outside Russia, as described in the Privacy Policy. The app shows the token count and the amount charged for every request. Server-side speech transcription is charged to your balance per minute at the price shown in the app; local transcription on your device is free.

Example

The cost of a request to a paid model is calculated as follows:

Take a typical agent task rather than a light one: 40 steps, each re-reading about 50,000 tokens of context (project code, conversation history, tool results) and writing about 1,000 tokens. That is 2,000,000 input and 40,000 output tokens in total. On Claude Sonnet 5 this task costs:

On the plan models the same task costs nothing on top of your subscription: the Pro plan allowance covers about 14 such tasks a month.

Compared with buying API access directly

The same task at the providers' public prices for models comparable to the plan models. For Krokki, the Pro subscription price is divided by the number of such tasks in its monthly allowance. Foreign prices are converted at the Bank of Russia rate on 24 Sep 2026 (₽84.40 per US dollar), exclude VAT and caching discounts, and cannot be paid directly with a Russian card. Yandex and Sber prices include VAT. Prices per 1 million tokens.

* Provider prices and the dollar rate are as of 24 Sep 2026 and are not updated on this page in real time, so current prices may differ. Check the providers' websites for up-to-date prices.

Topping up your balance

The balance pays for paid models. On the Pro and Max plans you can top it up in the app with ₽300, ₽1,000 or ₽3,000 by bank card, via the Russian Faster Payments System (SBP) or with other methods offered by the Robokassa payment service. A receipt is emailed to you after payment. Any unused balance can be refunded on request; see Refunds for details.

Prices may change. New prices apply to requests sent after they are published on this page; amounts already charged are not recalculated. A new subscription price applies from the next renewal and is published here at least 7 days in advance.