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

# Subagent server tool

> Delegate a self-contained task to a configured worker model during a response.

The `orq:subagent` tool delegates a self-contained task to another model. The worker receives the task, optional context, and the configuration on the tool entry. It does not receive the parent conversation.

Use a smaller or faster worker for extraction, summarization, drafting, or data transformation while the primary model continues to own the final response.

## Quick start

The examples use the client configuration from the [Server tools overview](/docs/ai-gateway/features/server-tools.md).

```bash cURL theme={"theme":{"light":"github-light","dark":"github-dark"}}
curl -X POST https://router.toqan.ai/eu/v3/router/chat/completions \
  -H "Authorization: Bearer $TOQAN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-5.4",
    "messages": [
      { "role": "user", "content": "Extract the dates and owners from this project update, then summarize the risks." }
    ],
    "tools": [
      {
        "type": "orq:subagent",
        "model": "openai/gpt-5.4-mini",
        "output_format": "JSON with dates, owners, and risks",
        "max_uses": 2
      }
    ]
  }'
```

```typescript TypeScript theme={"theme":{"light":"github-light","dark":"github-dark"}}
const response = await client.chat.completions.create({
  model: 'openai/gpt-5.4',
  messages: [
    {
      role: 'user',
      content:
        'Extract the dates and owners from this project update, then summarize the risks.',
    },
  ],
  tools: [
    {
      type: 'orq:subagent',
      model: 'openai/gpt-5.4-mini',
      output_format: 'JSON with dates, owners, and risks',
      max_uses: 2,
    },
  ] as any,
});

console.log(response.choices[0].message.content);
```

```python Python theme={"theme":{"light":"github-light","dark":"github-dark"}}
response = client.chat.completions.create(
    model="openai/gpt-5.4",
    messages=[
        {
            "role": "user",
            "content": "Extract the dates and owners from this project update, then summarize the risks.",
        }
    ],
    tools=[
        {
            "type": "orq:subagent",
            "model": "openai/gpt-5.4-mini",
            "output_format": "JSON with dates, owners, and risks",
            "max_uses": 2,
        }
    ],
)

print(response.choices[0].message.content)
```

The primary model supplies a `task` and optional `context` when it calls the tool. The worker's completed output is returned to the primary model.

## Configuration

| Parameter | Type | Required | Default | Description |
| - | - | - | - | - |
| `type` | string | Yes | | Must be `orq:subagent`. |
| `model` | string | Yes | | Worker model in `provider/model` format. |
| `max_tokens` | integer | No | Provider default | Maximum worker output tokens. Accepted range: 0 to 128,000, capped by the worker model's maximum output. `0` uses the provider default. |
| `reasoning_effort` | string | No | Provider default | `none`, `minimal`, `low`, `medium`, `high`, `xhigh`, or `max`, when supported by the worker model. |
| `temperature` | number | No | Provider default | Worker sampling temperature from 0 to 2. The selected model may impose a lower maximum. |
| `system_prompt` | string | No | Built-in worker prompt | Replace the worker's system prompt. |
| `output_format` | string | No | None | Add output-format guidance to the worker task. |
| `max_uses` | integer | No | Unlimited | Maximum delegations during the request. Set `0` or omit the field for no tool-specific limit. |

## Cost and usage

Worker tokens are billed at the selected model's standard rate. Each Subagent call appears at `usage.server_tool_use.subagent_requests`, including calls rejected by `max_uses` and calls whose worker request failed.

`orq:sidekick` remains accepted as a legacy alias.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.