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

# Built-in Agents

> Framework-level agents with deep platform integration. Configure only - cannot modify source.

<Note type="warning">
  **Built-in** - These agents are part of the Conductor framework itself. You can configure them, but cannot modify their source code.
</Note>

## What Are Built-in Agents?

Built-in agents are framework-level components that require deep integration with Cloudflare's platform:

* **Durable Objects** for stateful operations
* **Cloudflare AI** for embeddings and inference
* **Vectorize** for vector storage and search

Unlike [starter kit agents](/conductor/starter-kit/overview) which you own and can modify, built-in agents are maintained by the Conductor team and updated with framework releases.

## Available Built-in Agents

<CardGroup cols={2}>
  <Card title="rag" icon="database" href="/conductor/starter-kit/built-in/rag">
    Vector search + LLM using Cloudflare AI & Vectorize
  </Card>

  <Card title="hitl" icon="user-check" href="/conductor/starter-kit/built-in/hitl">
    Human-in-the-loop approval flows using Durable Objects
  </Card>
</CardGroup>

## Built-in vs Starter Kit

| Aspect            | Built-in Agents          | Starter Kit Agents            |
| ----------------- | ------------------------ | ----------------------------- |
| **Source code**   | Framework (not editable) | Your project (fully editable) |
| **Updates**       | With Conductor releases  | You control                   |
| **Platform deps** | Deep (DO, AI, Vectorize) | Minimal                       |
| **Customization** | Config only              | Full code access              |
| **Examples**      | rag, hitl                | fetch, validate, scrape, etc. |

## When to Use Built-in Agents

**Use `rag` when you need:**

* Manual control over document indexing
* Custom chunking strategies
* Hybrid search (vector + keyword)
* Fine-tuned retrieval parameters

**Use `hitl` when you need:**

* Human approval workflows
* Asynchronous task completion
* Audit trails for decisions
* Multi-step approval chains

## Configuration Pattern

Built-in agents are configured, not coded:

```yaml theme={null}
agents:
  - name: search-docs
    agent: rag
    config:
      vectorBinding: DOCS_VECTORS
      aiBinding: AI
      model: '@cf/baai/bge-base-en-v1.5'
      topK: 5
    input:
      query: ${input.question}
      operation: search
```

## Starter Kit Alternatives

For simpler use cases, consider starter kit agents:

| Need                 | Built-in | Starter Kit Alternative                     |
| -------------------- | -------- | ------------------------------------------- |
| RAG with zero config | -        | [autorag](/conductor/starter-kit/autorag)   |
| Simple validation    | -        | [validate](/conductor/starter-kit/validate) |
| HTTP requests        | -        | [fetch](/conductor/starter-kit/fetch)       |

## Next Steps

<CardGroup cols={2}>
  <Card title="RAG Agent" icon="database" href="/conductor/starter-kit/built-in/rag">
    Vector search + LLM
  </Card>

  <Card title="HITL Agent" icon="user-check" href="/conductor/starter-kit/built-in/hitl">
    Human approval flows
  </Card>

  <Card title="Utility Agents" icon="wrench" href="/conductor/starter-kit/fetch">
    Editable agents you own
  </Card>

  <Card title="AutoRAG" icon="magnifying-glass" href="/conductor/starter-kit/autorag">
    Zero-config managed RAG
  </Card>
</CardGroup>