> ## Documentation Index
> Fetch the complete documentation index at: https://docs.termique.app/llms.txt
> Use this file to discover all available pages before exploring further.

# AI assistant

> Get shell command suggestions and debug connection issues in any terminal session.

The AI assistant is a chat panel available inside any terminal session. Ask it to generate commands, explain output, or diagnose connection problems.

<Note>
  For an issue spanning multiple hosts, see [Incident investigation](/incident-investigation) instead.
</Note>

## Open the assistant

Click the **AI** icon in the terminal session toolbar. The chat panel opens alongside the terminal.

## What you can do

* **Generate commands**: describe what you want to do; the assistant returns a shell command you can run with one click.
* **Explain output**: paste terminal output and ask what it means.
* **Debug issues**: describe a connection error or unexpected behavior.
* **Automate tasks**: the assistant can propose a sequence of commands for multi-step operations.

## Provider modes

### Hosted (default)

By default, requests go to the Termique API backend. No API key required. Your conversation is processed on the backend and not stored.

Hosted usage has a daily limit: 15 prompts/day on Free, 100 prompts/day on [Pro](/plans). The limit resets at midnight UTC.

### BYO endpoint

If you have an OpenAI-compatible endpoint (e.g. your own OpenAI account, Ollama, or another provider), you can point the assistant at it.

1. Open **Settings → AI**.
2. Enable **Custom endpoint**.
3. Enter your endpoint URL and API key.

The API key is stored encrypted (using your DEK) and is only sent to your specified endpoint, never to the Termique backend.

<Warning>
  API keys are stored encrypted locally and in your API backend as an encrypted blob. Do not share your API key or enter it into an untrusted build of Termique.
</Warning>

## Auto-execute

The assistant can be configured to automatically run proposed commands without a manual confirmation click. This is off by default. Enable it in **Settings → AI → Auto-execute**.

Use auto-execute only on non-production hosts. Commands run immediately. There is no undo.

## Context

The assistant has context about which host you are connected to (label, hostname, username). It does not have access to your terminal history unless you paste it into the chat.

## Memory

On [Pro](/plans), the assistant can save memories (short notes about a host or your preferences) and recall them in later sessions. Free accounts don't have memory: each conversation starts fresh.