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

# Runtimes

> Choose and configure the execution engine that powers your agents. From general-purpose Agno to specialized Claude Code, select the runtime that matches your use case.

Runtimes are the execution engines that power your Kubiya agents. They determine how your agents process requests, interact with tools, manage conversations, and integrate with different LLM providers. Choosing the right runtime ensures optimal performance, cost efficiency, and capabilities for your specific use case.

<CardGroup cols={3}>
  <Card title="Agno Runtime" icon="zap" color="#c084fc">
    Multi-model runtime with support for all major LLM providers
  </Card>

  <Card title="Claude Code Runtime" icon="code" color="#7c3aed">
    Code-specialized runtime optimized for development workflows
  </Card>

  <Card title="Custom Runtimes" icon="puzzle-piece" color="#9333ea">
    Extend with your own framework (LangChain, CrewAI, AutoGen)
  </Card>
</CardGroup>

***

## What Are Runtimes?

A **runtime** is the execution layer that sits between your agent's configuration and the underlying AI model. It handles:

* **Model Integration**: Routing requests to different LLM providers (OpenAI, Anthropic, Google, etc.)
* **Tool Execution**: Managing Skills and MCP servers for agent capabilities
* **Conversation Management**: Maintaining context and history across multi-turn interactions
* **Streaming & Feedback**: Providing real-time execution updates
* **Performance Optimization**: Caching, batching, and efficient token usage

Kubiya's **runtime-agnostic architecture** lets you choose the execution engine that best fits your needs - whether it's a built-in runtime or your own custom implementation.

***

## Quick Comparison

| Feature            | Agno                  | Claude Code           |
| ------------------ | --------------------- | --------------------- |
| **Framework**      | Agno + LiteLLM        | Claude Code SDK       |
| **Model Support**  | All providers         | Claude only           |
| **Best For**       | Multi-model workflows | Code & development    |
| **Max History**    | 100 messages          | 200 messages          |
| **Specialization** | Provider flexibility  | Development-optimized |

<Card title="Full Feature Comparison" icon="scale-balanced" href="/core-concepts/runtimes/comparison">
  See detailed side-by-side comparison of all capabilities
</Card>

***

## Get Started

<Steps>
  <Step title="Choose Your Runtime">
    Determine which runtime best fits your use case:

    * **Multi-model flexibility needed?** → Agno Runtime
    * **Code generation & analysis?** → Claude Code Runtime
    * **Specialized framework (LangChain, CrewAI)?** → Custom Runtime

    <Card title="Compare Runtimes" icon="arrow-right" href="/core-concepts/runtimes/comparison">
      Use our decision matrix to choose the right runtime
    </Card>
  </Step>

  <Step title="Configure Your Agent">
    Select the runtime when creating an agent:

    **Via CLI:**

    ```bash theme={null}
    kubiya agent create my-agent \
      --runtime agno \
      --model kubiya/claude-sonnet-4
    ```

    **Via API:**

    ```bash theme={null}
    curl -X POST https://control-plane.kubiya.ai/api/v1/agents \
      -H "Authorization: Bearer $KUBIYA_API_KEY" \
      -H "Content-Type: application/json" \
      -d '{
        "name": "my-agent",
        "runtime": "agno",
        "model_id": "kubiya/claude-sonnet-4"
      }'
    ```

    <Card title="Agent Configuration" icon="arrow-right" href="/core-concepts/agents">
      Learn about agent creation and configuration
    </Card>
  </Step>

  <Step title="Execute and Monitor">
    Your agent will execute using the selected runtime. Monitor performance, token usage, and tool execution through the Kubiya dashboard.

    <Card title="Analytics & Monitoring" icon="arrow-right" href="/core-concepts/analytics/analytics">
      Track runtime performance and optimize costs
    </Card>
  </Step>
</Steps>

***

## Built-in Runtimes

<CardGroup cols={2}>
  <Card title="Agno Runtime" icon="zap" href="/core-concepts/runtimes/agno-runtime">
    **Multi-model flexibility** via LiteLLM

    * All LLM providers (GPT, Claude, Gemini, Mistral, Cohere)
    * Python-based tool integration
    * MCP server support
    * 100-message conversation history
    * Flexible execution engine

    **Best for**: Multi-model workflows, provider flexibility, general automation
  </Card>

  <Card title="Claude Code Runtime" icon="code" href="/core-concepts/runtimes/claude-code-runtime">
    **Code-optimized** execution

    * Claude models only (optimized integration)
    * Advanced file operations
    * Repository analysis capabilities
    * 200-message extended history
    * Session resumption for multi-turn

    **Best for**: Code generation, refactoring, development workflows, technical documentation
  </Card>
</CardGroup>

***

## Extend with Custom Runtimes

Need specialized capabilities? Build your own runtime using popular frameworks:

<AccordionGroup>
  <Accordion title="LangChain Integration" icon="link">
    Integrate LangChain's ecosystem of tools, chains, and agents. Build custom runtimes that leverage LangChain's composability while using Kubiya's orchestration layer.
  </Accordion>

  <Accordion title="CrewAI Multi-Agent Systems" icon="users">
    Create runtimes that coordinate multiple specialized agents using CrewAI's role-based architecture. Perfect for complex workflows requiring agent collaboration.
  </Accordion>

  <Accordion title="AutoGen Conversations" icon="comments">
    Build runtimes using Microsoft's AutoGen framework for advanced multi-agent conversations, code execution, and iterative problem-solving.
  </Accordion>

  <Accordion title="Custom Framework" icon="wrench">
    Implement the BaseRuntime interface to integrate any Python-based AI framework. Full control over execution logic, tool integration, and model interaction.
  </Accordion>
</AccordionGroup>

<Card title="Custom Runtime Developer Guide" icon="arrow-right" href="/core-concepts/runtimes/custom-runtimes">
  Complete guide with code examples, patterns, and best practices
</Card>

***

## Explore Documentation

<CardGroup cols={2}>
  <Card title="Understanding Runtimes" icon="book-open" href="/core-concepts/runtimes/overview">
    Learn how runtimes work and fit into the Kubiya architecture
  </Card>

  <Card title="Runtime Comparison" icon="scale-balanced" href="/core-concepts/runtimes/comparison">
    Side-by-side feature comparison and decision framework
  </Card>

  <Card title="Custom Runtimes" icon="puzzle-piece" href="/core-concepts/runtimes/custom-runtimes">
    Build your own runtime with custom frameworks
  </Card>

  <Card title="Control Plane" icon="network-wired" href="/core-concepts/control-plane/overview">
    Runtime registry and orchestration
  </Card>
</CardGroup>

***

## Related Concepts

<CardGroup cols={3}>
  <Card title="Agents" icon="robot" href="/core-concepts/agents">
    Configure agents with runtime selection
  </Card>

  <Card title="Skills" icon="layer-group" href="/core-concepts/skills">
    Tools and capabilities available to runtimes
  </Card>

  <Card title="Control Plane" icon="network-wired" href="/core-concepts/control-plane/overview">
    Runtime registry and orchestration
  </Card>
</CardGroup>