# Agent Playtest

an LLM agent drives a running SimVX game, headless.

```{raw} html
<span class="example-docs-badge">📄 Docs only</span>


```

**Tags:** `ai`

Run it two ways:

  # Real LLM loop against your OpenWebUI / vLLM / llama.cpp endpoint:
  SIMVX_LLM_BASE_URL=http://host:8000/v1 SIMVX_LLM_MODEL=your-model   SIMVX_LLM_API_KEY=sk-... uv run python examples/features/ai/agent_playtest.py

  # Offline (no endpoint): runs a scripted walk-through of the session verbs.
  uv run python examples/features/ai/agent_playtest.py

The game is a trivial "reach 5 points" toy so the loop is easy to follow; the
session verbs (observe / send_input / set_state / step, terminated vs
truncated) are identical for any real game.

## Source

```{literalinclude} ../../examples/features/ai/agent_playtest.py
:language: python
:linenos:
```