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

# Quick and Deep modes

> Two answering modes — a fast single-pass answer or thorough multi-step reasoning — chosen per message.

Every chat message is answered in one of two modes. **Quick mode** gives a fast, single-pass answer that's ideal for straightforward lookups. **Deep mode** works through a question in multiple steps — searching repeatedly, running data queries, building charts, using tools, and delegating to other agents when needed. You choose the mode per message, so you can move quickly on simple questions and go deep when a task calls for it.

## Quick mode

A single, fast pass over your knowledge base. Quick mode retrieves the most relevant passages and answers directly — perfect for definitions, policy lookups, and other straightforward questions where one good search is enough.

## Deep mode

Thorough, multi-step reasoning. Deep mode can search your documents more than once, run queries against connected databases, chart the results, call connected apps and tools, and delegate to other [agents](/agents/overview) to cover ground outside its own knowledge base. It's built for research, multi-part questions, and anything that needs data or tools to answer well.

Images and manually attached skills are handled in Deep mode.

<Info>
  [Agents](/agents/overview) always run in Deep mode, so agent runs get the full multi-step reasoning, tools, and delegation by default.
</Info>

## Choosing a mode

|                            | Quick mode               | Deep mode                     |
| -------------------------- | ------------------------ | ----------------------------- |
| Best for                   | Simple, direct questions | Research and multi-part tasks |
| Document search            | Single pass              | Repeated as needed            |
| Data queries and charts    | —                        | Yes                           |
| Tools and connected apps   | —                        | Yes                           |
| Delegating to other agents | —                        | Yes                           |
| Images and attached skills | —                        | Yes                           |
| Speed                      | Fastest                  | Thorough                      |

<Tip>
  Start with **Quick mode** for simple Q\&A — it's the fastest way to a grounded answer. Reach for **Deep mode** when you're doing research, asking a multi-part question, or need tools, data, or charts to get a complete answer.
</Tip>

## Tuning each mode

Persona and instructions are shared across Quick and Deep. Each mode also has its own full custom prompt, so you can replace the assembled prompt separately for Quick or Deep when needed.

<Card title="Customize prompts per mode" icon="sliders" href="/knowledge-bases/prompts">
  Set shared persona and instructions, plus mode-specific full custom prompts for Quick and Deep.
</Card>

## Next steps

<CardGroup cols={2}>
  <Card title="Chat overview" icon="comments" href="/chat/overview">
    How grounded, cited answers come together.
  </Card>

  <Card title="Streaming responses" icon="bolt" href="/chat/streaming">
    Watch answers arrive live, step by step.
  </Card>

  <Card title="Build an agent" icon="robot" href="/agents/overview">
    Agents run in Deep mode with tools and delegation.
  </Card>

  <Card title="Prompt customization" icon="sliders" href="/knowledge-bases/prompts">
    Shape each mode's voice and behavior.
  </Card>
</CardGroup>