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

# How Chamelio Orchestrates AI Models

> How Chamelio breaks a request into steps and assigns the right models and tools to each one

Chamelio works as an orchestration layer rather than relying on a single AI model. When you submit a request, Chamelio evaluates what the task actually requires and dynamically assigns the right models and tools for each step.

## How a request is handled

<Steps>
  <Step title="Decomposition">
    Complex requests are broken into smaller, well-defined sub-tasks. For example, a research request can be split into source discovery, filtering, extraction, and comparison.
  </Step>

  <Step title="Planning">
    Chamelio determines the most effective way to execute each sub-task, including which model to use, which data sources to draw on, and the order of the steps.
  </Step>

  <Step title="Execution">
    Each step runs with the models and tools best suited to it. This may involve querying external sources, analyzing documents, or applying legal reasoning.
  </Step>

  <Step title="Synthesis">
    The outputs of all steps are consolidated into a single, coherent response.
  </Step>
</Steps>

## Why it matters

<Columns cols={3}>
  <Card title="Accuracy and consistency" icon="bullseye">
    Smaller, well-defined steps reduce compounding errors on complex requests
  </Card>

  <Card title="Multi-source analysis" icon="layer-group">
    Combine information from your documents, the conversation, and external sources reliably
  </Card>

  <Card title="Transparency" icon="eye">
    Reasoning and planned actions are exposed where applicable, so you can see how conclusions were reached
  </Card>
</Columns>

The result is AI-powered legal work that is both powerful and predictable, without sacrificing control or clarity.

## Related topics

<Columns cols={2}>
  <Card title="Web Agent Context" icon="brain" href="/features/core/agent-web-context">
    How the context window shapes what the Web Agent can evaluate at once
  </Card>

  <Card title="Agent Advanced Features" icon="sparkles" href="/features/core/agent-advanced-features">
    Multi-modal input, intelligent processing, and output formats
  </Card>
</Columns>