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

# Reasoning

> Enable extended thinking with WithLLMSampling and stream provider-specific reasoning deltas

Configures generic `LLMReasoning` through `WithLLMSampling` and streams the run to observe thinking output (e.g. Anthropic extended thinking).

Source: [`examples/agent_with_reasoning/`](https://github.com/agenticenv/agent-sdk-go/tree/main/examples/agent_with_reasoning)

## What it demonstrates

* `WithLLMSampling` with `Reasoning: &interfaces.LLMReasoning{...}`
* Provider-specific mapping of `Enabled`, `Effort`, `BudgetTokens`
* Streaming reasoning-related events

## Run

From `examples/`:

```bash theme={null}
go run ./agent_with_reasoning
go run ./agent_with_reasoning "Why is the sky blue? One short paragraph."
```

Use a model that supports extended thinking for your provider — for example `claude-sonnet-4-20250514` with Anthropic, or `o3` with OpenAI.

## Expected output

With Anthropic extended thinking enabled you will see thinking deltas on stderr (or as events) followed by the final answer:

```
[thinking] The blue color of the sky is due to a phenomenon called Rayleigh scattering...
The sky appears blue because of Rayleigh scattering — shorter blue wavelengths scatter
more than longer red ones as sunlight passes through the atmosphere.
```

Thinking content appears only when the provider emits it; output format varies by provider.

## Learn more

<CardGroup cols={2}>
  <Card title="Reasoning" icon="lightbulb" href="/features/reasoning">
    Field mapping per provider
  </Card>

  <Card title="Stream" icon="wave-pulse" href="/examples/stream">
    Streaming event handling
  </Card>
</CardGroup>