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

# Recordings & Replays

> Viewing your agent's gameplay

Recordings let you view and share your agent's gameplay sessions.

## When Recordings Are Available

| Method                   | Recordings Available                                       |
| ------------------------ | ---------------------------------------------------------- |
| **API**                  | Yes — viewable online via scorecard                        |
| **Benchmarking harness** | Yes — saved locally and viewable online                    |
| **Python Toolkit**       | Optional — set `save_recording=True` when calling `make()` |

## Online Replays

For games played via the API, you can view recordings online through your scorecard:

`https://arcprize.org/scorecards/<scorecard_id>`

Here is an example [recording](https://arcprize.org/replay/1d251d20-9043-4ace-9f9d-09822f5438d8).

## Local Recording Files

The Python Toolkit saves local or remote runs when you enable recording:

```python theme={null}
env = arc.make(
    "ls20",
    save_recording=True,
    include_frame_data=True,
)
```

Files are written beneath the configured `recordings_dir` using this layout:

```text theme={null}
recordings/<scorecard_id>/<game_id>-<guid>.jsonl
```

Set `include_frame_data=False` to omit frame arrays and reduce file size. The [benchmarking harness](/benchmarking-agent) also writes run recordings, model responses, token usage, and per-step records beneath `recordings/`.

## Recording File Format

### JSONL Format

Recordings are stored in JSONL format with timestamped entries:

```json theme={null}
{"timestamp": "2026-07-22T10:30:45.123456+00:00", "data": {"game_id": "ls20-016295f7601e", "state": "NOT_FINISHED", "levels_completed": 0, "win_levels": 7, "action_input": {"id": "RESET", "data": {}, "reasoning": null}, "guid": "...", "full_reset": true, "available_actions": [1, 2, 3, 4, 6], "frame": [...]}}
{"timestamp": "2026-07-22T10:30:46.234567+00:00", "data": {"game_id": "ls20-016295f7601e", "state": "NOT_FINISHED", "levels_completed": 0, "win_levels": 7, "action_input": {"id": "ACTION1", "data": {}, "reasoning": {"thought": "move up"}}, "guid": "...", "full_reset": false, "available_actions": [1, 2, 3, 4, 6], "frame": [...]}}
```