> For the complete documentation index, see [llms.txt](https://docs.langx.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.langx.io/docs/library/language-copilot.md).

# LangX Copilot

LangX Copilot, the AI helper in a chat: what it does, what it sends, and why it is a beta that may not be on for your account yet.

LangX Copilot is an AI helper inside a chat with a real person. It helps you write and understand messages; it never talks for you, and nothing it writes reaches the other person unless you send it yourself.

## Beta, rolling out

Copilot is part of **Pro** and is in **beta, rolling out**: it is being turned on for accounts gradually, so it may not be on for yours yet, even on Pro. When it is not, the app simply shows no Copilot button.

It is never paid for with tokens: [tokens](/docs/token/utility.md) are spent on a streak freeze and on cosmetics, and they never unlock a feature.

## What it does

* **Check my sentence.** Before you send, it looks over what you wrote and suggests a better version, or tells you it looks good.
* **Say it in the language you are learning.** Write what you mean and it suggests how to say it.
* **Explain.** Hold the other person's message, or a correction, and choose Explain to see what it means and why.
* **Reply ideas.** When the other person's newest message is waiting for an answer, a chip above your message box offers a few short replies at your level, for you to edit.

Every answer carries a blue **Copilot · Beta** label, so it is never mistaken for a correction from a person, which is green. Tapping a suggestion puts it in your message box; it does not send it.

## Before you use it

The first time, the app asks before anything is sent. When you ask Copilot something, the text you asked about goes to Anthropic, the company that runs the AI model, with your languages and levels. To explain a message, that is the one message you picked, which the other person wrote. For reply ideas, it includes the last few messages in the chat. Your name, email, photos and voice notes are never sent. AI can be wrong, so read an answer before you use it.

**Settings → Privacy → Copilot** turns it off again: nothing more of yours is sent, and Copilot asks before it is used the next time. Every answer also has **Report**, which sends that answer to the LangX team. The [privacy policy](https://langx.io/privacy-policy) (section 2.6) says exactly what each help sends and what is kept.

There is a daily number of answers, and the app shows how many are left today.