Cursor in Russian: interface, AI replies and code comments

Three different questions behind one query: the language of AI replies, of the interface and of code comments. Each covered, with .cursor/rules and a token calculation.

6 min · · IDE · development

Cursor AI in Russian is a query that hides three different questions: in what language the AI replies, in what language the editor interface is shown, and in what language to write code comments. They are solved differently. You fully control the reply language with rules. The interface is less certain, and below we say honestly what we checked and what you need to look at yourself. A separate section covers money: Russian text uses tokens differently from English.

The language of AI replies: project rules

Models understand requests in Russian and reply in it, but by default the agent may answer in English, especially if the surrounding code and comments are English. A reliable way is to fix the language in rules.

According to the Cursor documentation, project rules live in .cursor/rules as files with the .mdc extension, with a metadata block at the top containing the fields description, globs and alwaysApply. With alwaysApply: true the rule is always attached. An example .cursor/rules/language.mdc:

bash
---
description: Language of communication and comments
alwaysApply: true
---
- Reply in Russian.
- Write code comments and docstrings in Russian.
- Keep identifiers (variables, functions, classes) and commit messages in English.
- Do not translate technical terms (middleware, endpoint, rate limit).

For simple cases there is an alternative: an AGENTS.md file at the project root, plain markdown without metadata. The Cursor documentation calls it an option for simple instructions. If you want Russian to apply in all projects, look in Cursor's settings for the user rules section (the documentation calls it User Rules) and add the same line "Reply in Russian". How to write rules and not waste tokens on them is covered in the article on Cursor rules.

A wording tip: write the rule short and unambiguous. "Reply in Russian" works better than a paragraph about how you like to communicate. And do not mix requirements: list separately what must be in Russian (replies, comments) and what must not (identifiers, commits, file names).

The interface: what we know and what we do not

Care is needed here. What can be stated:

  • Cursor is an editor based on VS Code, and many users ask whether a Russian language pack can be installed, as in VS Code.
  • The Cursor documentation site has a language switcher, and among the options is Russian, so documentation pages can be read in Russian.
  • In the documentation pages about settings and extensions that we looked through while preparing this article, we found no separate interface language setting.

What we do not claim: that the whole Cursor interface, including its own chat, agent and settings panels, is localized into Russian, or that a language pack installs without quirks. If you need a Russian interface, check yourself: open the extensions panel, look for a Russian language pack and see whether it appears in search and what gets translated after installation. Some of Cursor's own screens may stay in English. A practical tip: treat an English interface as a given and set the working language (replies, comments) with rules.

Comments, commits and documentation in code

Decide in advance what is in Russian and what is not. A typical agreement:

WhatLanguageWhy
Agent replies in chatRussianyou read them
Comments and docstringsRussian or English, per teamdepends on who reads the code
Variable and function namesEnglishthe standard, and compatible with libraries
Commit messagesper the project's conventionto keep history uniform
README and documentationthe audience's languagewho it was written for

If the team includes people who do not read Russian, keep comments in English and let chat replies be in Russian for you personally: this is configured separately, with a personal rule and a project rule.

How Russian text affects tokens

You pay for both input and output tokens. Russian text usually takes more tokens than English for the same meaning; how much depends on the model and tokenizer, so a multiplier cannot be named in advance, but it is easy to measure. The exact number for your text comes from a count_tokens request in the Anthropic format (documentation):

bash
curl https://api.zukko.pro/v1/messages/count_tokens \
  -H "Authorization: Bearer fetch_your_key" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-sonnet-5-5",
    "messages": [{"role": "user", "content": "Отвечай на русском языке. Комментарии пиши на русском."}]
  }'

The response has an input_tokens field. Run the Russian and English versions of your rules file this way and compare. More on the reasons is in the article why Russian text is counted differently.

What this means in money. Assumption for the example: the Russian rules take 300 more tokens than the English version (measure your own). If Cursor is connected to Zukko in the OpenAI format, the rules go into the input of every request, and all Claude input in that format is billed at the input price. Formula: extra tokens × requests × input price.

ModelCalculation per 1,000 requests
Claude Sonnet 5 (named cursor-sonnet-5 in Cursor)300 × 1,000 = 0.3M tokens × $0.4 = $0.12
gpt-6.1-sol0.3M × $0.27 = $0.081
deepseek-v4.1-flash0.3M × $0.013 = $0.0039

Prices as of 18 October 2026. The sum is small, but it grows with the length of the rules. Hence the advice: one short rule about language plus other rules only where needed (via globs), not ten pages attached always.

Replies in Russian are also more expensive in output tokens, and output costs more than input (for claude-sonnet-5-5 $2 versus $0.4 per million). If you do not need lengthy explanations, add "reply briefly" to the rule: it cuts both time and spending.

Connecting Cursor to Zukko

If you want to use Cursor with your own key, you need a fetch_... key from the dashboard and the address https://api.zukko.pro/v1. Cursor's model settings have a place for your own OpenAI key and base address; exact item names change between versions, so follow the Cursor documentation. Take model names from the table in that documentation: Cursor will not let you add a model under a name it already knows, so the gateway has second names for some models (for example, cursor-sonnet-5 for Claude Sonnet 5). Check Cursor's own subscription terms on the service's site. The Zukko balance can be topped up via SBP, MasterCard or crypto.

FAQ

Does Cursor have a Russian language?

You can get AI replies in Russian, and it is configured with rules. The Cursor documentation is also available in Russian. For full interface localization we found no unambiguous confirmation: check in your version whether a Russian language pack is available.

How do I make Cursor always reply in Russian?

Add a rule with alwaysApply: true in .cursor/rules, or a line "Reply in Russian" in user rules so it applies in all projects. You can also use AGENTS.md at the project root.

Is Russian more expensive than English?

Usually Russian text takes more tokens than English for the same meaning. Measure the difference for your text with count_tokens. In practice, keep rules short and ask for brief replies.

Should code comments be in Russian or English?

The team decides. If only Russian-speaking colleagues read the code, Russian comments are convenient. Identifiers are better left in English in any case.

zukko

Pay 25×
less

for Claude Code & ChatGPT

One fetch_* key and base URL api.zukko.pro — for Cursor, Claude Code, Codex, and any SDK.

Get API keyDocumentation

api.zukko.pro · fetch_*

Cursor AI in Russian: replies, rules, interface · Zukko