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

# Agents

> LLM-powered assistants that reason across your data to answer questions and generate insights.

An **Agent** is an LLM-powered assistant that reasons across your room's data — its [spaces](/concepts/spaces), tables, and sources — to answer questions, run analysis, and generate insights. Every response is grounded in your verified enterprise data, with a transparent view of how the agent got there.

<Note>
  Agents combine retrieval over your embeddings with an LLM's reasoning, then orchestrate multi-step actions — querying tables, executing code, generating visualizations — to deliver context-aware answers.
</Note>

## How an agent works

1. **Understands** your question in natural language.
2. **Retrieves** relevant context from the room's spaces and data.
3. **Reasons and acts** — chaining steps like querying tables, running Python, or calling tools.
4. **Responds** with a grounded answer, including citations and a visible thought process.

## What shapes an agent

| Aspect               | Description                                                                                  |
| -------------------- | -------------------------------------------------------------------------------------------- |
| **Data & spaces**    | The embeddings and tables the agent reasons over.                                            |
| **Completion model** | The LLM that powers responses — choose from built-in models, **Auto**, or your own endpoint. |
| **Skills**           | Reusable capabilities that extend what the agent can do.                                     |
| **Tools**            | Built-in capabilities like web search, plus custom APIs the agent can call from code.        |

## Conversations, explainability, and sharing

Each conversation is a **thread** you can revisit, rename, and share as a read-only link. Every answer exposes its reasoning chain and citations, and finished agents can be deployed as branded public apps or connected over the MCP protocol.

<Card title="Agents (Feature Guide)" icon="robot" href="/features/agents">
  Configure models, chat, explainability, and sharing.
</Card>

## Example questions

```
Show me accounts with unusual transaction patterns
What are the common characteristics of high-value customers?
Find documents related to compliance requirements
```

## Related Concepts

<CardGroup cols={2}>
  <Card title="Spaces" icon="layer-group" href="/concepts/spaces">
    The embeddings agents reason over.
  </Card>

  <Card title="Sharing & Collaboration" icon="share-nodes" href="/features/sharing">
    Share threads and artifacts.
  </Card>

  <Card title="Integrations & MCP" icon="plug" href="/integrations/overview">
    Connect agents to your applications.
  </Card>

  <Card title="Deployment" icon="rocket" href="/deployment/overview">
    Publish agents as public apps.
  </Card>
</CardGroup>