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

# Limits, retention, and usage

> Plan for service limits and retention, and inspect the fields available to your account.

Use the values below for initial planning, then inspect the limits returned for the authenticated account.

| Boundary | Behavior |
| - | - |
| Simultaneous operations | 50 during Beta; check the account-effective limit and contact the team to request a higher limit |
| Checkpoint images | Retained for 180 days; after that, untagged and unreferenced images may be deleted |
| Recorded stdin, stdout, stderr | Spawn default: 1 MiB per stream; maximum: 10 MiB per stream |
| Execution timeout | Request field is seconds; account-effective maximum is exposed in token limits |
| Networking | Enabled by default; `networking.enabled=false` starts the VM without a guest network interface |

Contact [contree@nebius.com](mailto:contree@nebius.com) before a workload depends on file-upload, download, total-image, start-rate, regional, or event-retention limits. Also confirm any required deletion behavior, cancellation retention, or retention treatment for tagged and referenced images. Use the same contact to request a higher simultaneous-operation limit.

## Inspect effective limits

Token information exposes a `limits` map for the authenticated principal. Treat those values as account-effective controls, because they may differ from the overview. The `statistics` field is reserved for future use and is not a usage report.

Complete [Set up access](/sandboxes/start/set-up-access) and install `curl` and `jq`. Run both examples in the same shell with `NEBIUS_API_KEY` and `NEBIUS_PROJECT_ID` from access setup. Use the base URL below unless your deployment requires a different documented URL.

```bash theme={null}
CONTREE_BASE_URL="https://api.tokenfactory.nebius.com/sandboxes"

curl --max-time 20 --fail-with-body -sS "$CONTREE_BASE_URL/v1/whoami" \
  -H "Authorization: Bearer $NEBIUS_API_KEY" \
  -H "Project: $NEBIUS_PROJECT_ID" |
  jq '{permissions, limits}'
```

Each completed operation exposes raw resource fields including CPU, memory, I/O, process-switch, fault, signal, swap, `elapsed_time`, and `cost`. Use these fields for diagnostics only unless the Sandboxes team has confirmed their units and aggregation for the intended calculation. Use the operation-list `duration` field for elapsed seconds. Use billing records, rather than the operation `cost` field, for invoice reconciliation.

Replace `YOUR_OPERATION_UUID` with the accepted operation UUID from a completed run. [Run commands and handle results](/sandboxes/guides/run-commands-and-results#start-observe-and-finish) explains how to submit and observe one. Query its status, process state, captured streams, and raw resources:

```bash theme={null}
OPERATION_ID="YOUR_OPERATION_UUID"

curl --max-time 20 --fail-with-body -sS \
  "$CONTREE_BASE_URL/v1/operations/$OPERATION_ID" \
  -H "Authorization: Bearer $NEBIUS_API_KEY" \
  -H "Project: $NEBIUS_PROJECT_ID" |
  jq '{uuid,status,error,duration,image_uuid,result_image_uuid,
       process:.metadata.result.state,resources:.metadata.result.resources}'
```

Use `GET /v1/operations` for operation history. Before comparing workloads with resource metrics, confirm the units and aggregation of every selected field.

## Persistence consequences

A retained operation can return `result_image_uuid`; a disposable operation cannot provide a continuation point. See [Interpret the result image](/sandboxes/concepts/images-checkpoints-and-branches#interpret-the-result-image) for unchanged filesystems and interrupted or failed runs.

See [Troubleshooting](/sandboxes/operate/troubleshooting) and [Update images and clients](/sandboxes/operate/update-images-and-clients).