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

# Agent

>  Cerebrum is Ardor's built-in AI copilot that guides you through the platform. Whatever you can do in Ardor, Cerebrum can do as well.

<img src="https://mintcdn.com/ardor/sf-RwXk0pWSNetwj/docs/cerebrum/images/copilot.webp?fit=max&auto=format&n=sf-RwXk0pWSNetwj&q=85&s=67f884fa485fa2eb34517cb9b7174307" alt="Cerebrum Copilot" width="1303" height="831" data-path="docs/cerebrum/images/copilot.webp" />

## Referencing Context

<img src="https://mintcdn.com/ardor/sf-RwXk0pWSNetwj/docs/cerebrum/images/referencing-context.webp?fit=max&auto=format&n=sf-RwXk0pWSNetwj&q=85&s=02838c581efa4f62f1f1a3a364c53856" alt="Referencing Context" width="1317" height="715" data-path="docs/cerebrum/images/referencing-context.webp" />

In chat, use `@` to reference a solution its resources, an artifact or visual assets.

When you add a reference, that context is passed to Cerebrum with your message. This helps the agent focus on the exact part of Ardor you want to inspect, change, debug, or explain.

<Tip>
  Use references when your request depends on a specific object, such as "check logs for `@api-service`", "update `@landing-page`", or "explain what is inside `@my-solution`".
</Tip>

## Available Models

Cerebrum integrates with leading AI providers, giving you access to the most powerful models available:

| Model             | Provider  | Available From |
| ----------------- | --------- | -------------- |
| Gemini 3.1 Pro    | Google    | Free           |
| Gemini 3.5 Flash  | Google    | Free           |
| GPT-5.4           | OpenAI    | Starter+       |
| Grok 4.20         | xAI       | Starter+       |
| Claude Sonnet 4.6 | Anthropic | Starter+       |
| Claude Opus 4.6   | Anthropic | Pro only       |
| Claude Opus 4.7   | Anthropic | Pro only       |

### How to Switch Models

<img src="https://mintcdn.com/ardor/sf-RwXk0pWSNetwj/docs/cerebrum/images/model-selection.webp?fit=max&auto=format&n=sf-RwXk0pWSNetwj&q=85&s=8b83e4e3b8c1b0060bbd8e9756dfae3f" alt="Model selection" width="1131" height="816" data-path="docs/cerebrum/images/model-selection.webp" />

1. Click on the model name at the bottom of the chat
2. Select your preferred model from the dropdown
3. Continue your conversation with the new model

### Choosing the Right Model

<AccordionGroup>
  <Accordion title="Gemini 3.1 Pro" icon="google">
    **Best for:** Frontier reasoning, complex coding, UI work, and long-context synthesis

    A frontier-grade model available on the Free plan. Use it for serious Copilot work that needs reasoning, code understanding, multimodal context, or large-context synthesis.
  </Accordion>

  <Accordion title="Gemini 3.5 Flash" icon="zap">
    **Best for:** Frontier agentic coding, multimodal work, long-horizon tasks, and fast iteration

    A fast frontier model available on the Free plan. It pairs serious coding and agentic workflow capability with low relative burn and speed, making it suitable for real implementation work.
  </Accordion>

  <Accordion title="GPT-5.4" icon="microchip">
    **Best for:** Complex reasoning, advanced code generation, nuanced instructions

    OpenAI's professional model for difficult planning, debugging, and implementation tasks.
  </Accordion>

  <Accordion title="Grok 4.20" icon="bolt">
    **Best for:** Marketing copy, exploratory writing, and brainstorming

    xAI's model is useful for marketing-oriented text and broad ideation. Long contexts can consume credits faster, so use it when that extra context is useful.
  </Accordion>

  <Accordion title="Claude Sonnet 4.6" icon="feather">
    **Best for:** Long documents, detailed analysis, nuanced writing

    Anthropic's balanced model offering great performance at a reasonable cost. Excellent for code review and documentation.
  </Accordion>

  <Accordion title="Claude Opus 4.6" icon="gem">
    **Best for:** Anthropic frontier reasoning, long-running agents, and professional code review

    This Opus model is Pro-only because it is significantly more expensive to run. Choose it when Claude's reasoning, review style, or long-running agent behavior is the right fit.
  </Accordion>

  <Accordion title="Claude Opus 4.7" icon="gem">
    **Best for:** Anthropic frontier reasoning, long-running agents, and professional code review

    This Opus model is Pro-only because it is significantly more expensive to run. Choose it when Claude's reasoning, review style, or long-running agent behavior is the right fit.
  </Accordion>
</AccordionGroup>

## Working with Cerebrum

### Example Prompts

<CardGroup cols="2">
  <Card title="Agentic App" icon="robot">
    "Build a customer support chatbot that can answer questions about our product documentation"
  </Card>

  <Card title="Web App" icon="globe">
    "Create a todo app with user authentication and a PostgreSQL database"
  </Card>

  <Card title="API Service" icon="code">
    "Build a REST API for managing inventory with CRUD operations"
  </Card>

  <Card title="Data Pipeline" icon="diagram-project">
    "Set up a pipeline that fetches data from an API and stores it in a database"
  </Card>
</CardGroup>

## Best Practices

<AccordionGroup>
  <Accordion title="Be Specific">
    The more details you provide, the better Cerebrum can understand your requirements. Include information about expected users, data formats, and integration needs.
  </Accordion>

  <Accordion title="Spec-Driven Development">
    Ask Cerebrum to plan the changes you want to make, ask clarifying questions, and create documentation based on your answers. That way, you can see what the agent is going to do for you before it makes any changes.
  </Accordion>

  <Accordion title="Iterate">
    Don't try to describe everything at once. Start with core functionality and iterate with Cerebrum to add features.
  </Accordion>
</AccordionGroup>

## Getting Started

<Card title="Try Cerebrum Now" icon="rocket" href="https://ardor.cloud">
  Sign up for Ardor Cloud and start building with Cerebrum today. No credit card required for the free tier.
</Card>
