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

# AI assistant overview

> What the agent does, how the sidebar is laid out, and where to find the rest of the AI features.

Rumus's AI assistant is more than a chat box — it's an **agent** that can read files, run commands, search the web, call tools through MCP, and complete multi-step jobs end-to-end. This page gives you the lay of the land; deeper pages cover each feature in detail.

## Where it lives

The AI lives in the **right sidebar**. Toggle it from the panel-right icon at the far right of the title bar.

When the sidebar is open, you're talking to a single conversation. Open the **History** view (also in the sidebar) to switch between past conversations or start a new one.

## What it can do

<CardGroup cols={2}>
  <Card title="Answer questions" icon="message">
    Explain a command, summarize a log, sketch a fix.
  </Card>

  <Card title="Run commands" icon="play">
    Propose a command, ask for approval (when needed), execute it in your active terminal — see [Agentic execution](/docs/ai/agentic).
  </Card>

  <Card title="Plan multi-step work" icon="list-check">
    For larger jobs, the agent automatically drafts a plan and works through it — see [Plan mode](/docs/ai/plan-mode).
  </Card>

  <Card title="Search the web" icon="globe">
    Pull current information from the open web — see [Web search](/docs/ai/web-search).
  </Card>

  <Card title="Call your own tools" icon="plug">
    Hook up any MCP server (databases, internal APIs, ticket systems) — see [MCP integration](/docs/ai/mcp).
  </Card>

  <Card title="Follow your conventions" icon="book">
    Always-on rules and named skills the agent invokes when relevant — see [Rules & skills](/docs/ai/rules-skills).
  </Card>
</CardGroup>

## What you'll see in the sidebar

A typical AI message has more on it than just text:

* **Reasoning blocks** — collapsible "thinking" sections from reasoning models, when the model exposes them.
* **Tool calls** — when the agent runs a command, fetches a URL, or calls an MCP tool, you see the call inline with its result.
* **Plan blocks** — when the agent enters [plan mode](/docs/ai/plan-mode), the plan renders as a checklist with status icons.
* **Token info** — click the small info icon on a message to see exactly how many input / output / cached tokens that response used (and what it cost on built-in models).

## Picking a model

Click the model name above the input box to switch. The list combines:

* **Built-in models** included with your Rumus subscription — see [Built-in models](/docs/models/built-in).
* **Custom models** you've added with your own API key — see [Models & Providers](/docs/models/overview).

The model picker remembers your last choice **per conversation**, so you can keep parallel threads on different models.

## Smart autocomplete

Separate from the chat: as you type in any terminal tab, Rumus suggests context-aware completions inline. Press **→** or **Tab** to accept. See [Smart autocomplete](/docs/ai/autocomplete).

## Settings tour

The AI's behavior is configured at **Settings → AI**, split across four tabs:

| Tab              | What's in it                                                                        |
| ---------------- | ----------------------------------------------------------------------------------- |
| **General**      | Master enable/disable, conversation defaults, web search, browser engine            |
| **Models**       | Add and manage AI providers and models — see [Models & Providers](/docs/models/overview) |
| **Conversation** | Auto-naming, task lists, command approval, autocomplete tuning                      |
| **MCP**          | Custom MCP servers and their connection status                                      |

There's also a separate **Rules & Skills** entry at the same level — see [Rules & skills](/docs/ai/rules-skills).

## Privacy

* **Built-in model traffic** is routed through Rumus to the upstream provider. Rumus logs token counts but not message contents — see [Built-in models](/docs/models/built-in#privacy-of-built-in-model-traffic).
* **BYOK (your own API key) traffic** goes directly from your machine to the provider you configured. Rumus never sees the prompts.
* API keys, host credentials, and anything else sensitive live in the encrypted [vault](/docs/account/vault).

## Where to go next

<CardGroup cols={2}>
  <Card title="Agentic execution" icon="robot" href="/docs/ai/agentic">
    What "the agent runs commands" actually means in practice.
  </Card>

  <Card title="Plan mode" icon="list-check" href="/docs/ai/plan-mode">
    How and when plans get auto-generated.
  </Card>

  <Card title="Command approval" icon="shield" href="/docs/ai/command-approval">
    The whitelist / blacklist that gates what runs without asking.
  </Card>

  <Card title="Rules & skills" icon="book" href="/docs/ai/rules-skills">
    Teach the agent your conventions once and reuse them everywhere.
  </Card>
</CardGroup>