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

# Serverless Agents

> Deploy agents from the Workbench without managing code or infrastructure

Serverless is the default deployment mode when you build agents entirely in the **Workbench** without customizing code. Your agent runs on agent-workers (xpander cloud or self-hosted) without you managing containers or code files.

<Note>
  **When you customize agent code** (write your own `xpander_handler.py`), your agent automatically becomes containerized when you run `xpander deploy`. See [Agent Containers](/guides/run/agent-containers) for the code-first workflow.
</Note>

***

## How It Works

Build your agent visually in Workbench:

1. **Configure in UI**: Set LLM model, system prompt, tools, knowledge bases
2. **Click Deploy**: Agent runs on serverless agent-workers
3. **Invoke**: Use REST API, SDK, Slack, webhooks, or any integration

**No code required.** The platform handles everything.

**Agent-worker architecture:**

* **xpander Cloud**: Managed agent-workers that auto-scale
* **Self-Hosted**: Deploy agent-workers to your own Kubernetes cluster

<Steps>
  <Step title="Build Agent in Workbench">
    Configure your agent using the visual interface:

    * **LLM Settings**: Choose model (GPT-5, Claude Sonnet 4.5, etc.)
    * **Instructions**: Define agent role, goals, and behavior
    * **Connectors**: Add tools (Slack, GitHub, Google, etc.)
    * **Knowledge Bases**: Upload documents for RAG
    * **Memory**: Enable user memories and session storage
  </Step>

  <Step title="Deploy from Workbench">
    Click **Deploy** in Workbench.

    Your agent automatically runs on serverless agent-workers.

    No Dockerfile, no CLI commands, no code files needed.
  </Step>

  <Step title="Invoke from Anywhere">
    Call your agent via REST API, SDK, Slack, webhooks, or any integration:

    ```python theme={"dark"}
    from xpander_sdk import Backend, Configuration

    backend = Backend(configuration=Configuration(api_key="<your-key>"))

    result = backend.invoke_agent(
        agent_id="<agent-id>",
        input="Your message",
        user_id="user@example.com",
        session_id="session-123"
    )
    ```
  </Step>
</Steps>

***

## Benefits

<CardGroup cols={2}>
  <Card title="Zero Code" icon="wand-magic-sparkles">
    Build agents visually without writing code
  </Card>

  <Card title="Zero Infrastructure" icon="server">
    No containers, servers, or Kubernetes to manage
  </Card>

  <Card title="Instant Deployment" icon="rocket">
    Deploy from Workbench in seconds
  </Card>

  <Card title="Auto-Scaling" icon="arrows-up-down">
    Agent-workers automatically scale based on load
  </Card>
</CardGroup>

***

## Serverless vs Containers

| Feature            | Serverless (No Code)                   | Containers (Code-First)             |
| ------------------ | -------------------------------------- | ----------------------------------- |
| **How you build**  | Workbench UI only                      | Write `xpander_handler.py` code     |
| **Deployment**     | Click Deploy in Workbench              | Run `xpander deploy` CLI            |
| **Infrastructure** | Agent-workers (managed or self-hosted) | Docker containers + registry        |
| **Complexity**     | Zero code, zero configuration          | Write code, manage Dockerfile       |
| **Flexibility**    | Pre-configured runtime                 | Full control over environment       |
| **Use case**       | Most agents                            | Custom dependencies, advanced logic |

***

## Where Agent-Workers Run

### xpander Cloud (Default)

Your agent runs on xpander's managed agent-workers:

* Automatic scaling
* Pay-as-you-go pricing
* No infrastructure to manage
* Data processed in xpander cloud

### Self-Hosted

Deploy agent-workers to your own Kubernetes:

* Full data control
* Runs in your infrastructure
* Deploy once, use for all serverless agents
* See [Self-Hosted Deployment](/self-hosted/index)

***

## Configuration

### Environment Variables

Set environment variables in Workbench **Settings** → **Environment Variables**:

```bash theme={"dark"}
# Custom configuration
DATABASE_URL=postgresql://...
API_ENDPOINT=https://api.example.com
```

### Resource Limits

Configure in Workbench **Settings** → **Resources**:

* **CPU**: 0.25 - 4 vCPUs
* **Memory**: 512MB - 8GB
* **Timeout**: 30s - 15min

***

## Monitoring

View execution metrics in **Monitor** tab:

* Request count and latency
* Memory and CPU usage
* Error rates
* Token usage
* Thread history and logs

See [Observability](/guides/manage/observability) for detailed monitoring.

***

## When to Use

**Use Serverless (no code) when:**

* ✅ Building agents with Workbench UI
* ✅ Don't need custom code logic
* ✅ Standard connectors and tools are sufficient
* ✅ Want zero infrastructure management

**Use Containers (code-first) when:**

* ❌ Need to write custom agent logic
* ❌ Require custom system dependencies
* ❌ Need specific Python packages or base images
* ❌ Want full control over agent behavior

<Note>
  **Taking over agent code?** Once you customize the agent code with `xpander agent new`, your agent automatically becomes containerized when you run `xpander deploy`. See [Agent Containers](/guides/run/agent-containers).
</Note>

***

## Related

<CardGroup cols={2}>
  <Card title="Agent Containers" icon="box" href="/guides/run/agent-containers">
    Package agents as Docker containers
  </Card>

  <Card title="Self-Hosted Deployment" icon="server" href="/self-hosted/index">
    Run agent-workers on your infrastructure
  </Card>
</CardGroup>