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

# Instance lifecycle

> Payment, provisioning, active runtime, and irreversible shutdown.

## Payment and provisioning

An unpaid checkout is not a running GPU. After payment is confirmed, the deployment provisions. Wait for active status and access details before attempting SSH.

## Active runtime

The deployment's `startedAt` and `endsAt` fields describe its active runtime. API timestamps use UTC. Use the actual deployment end time rather than assuming the checkout time is the start.

## Expiration

The backend schedules shutdown when purchased runtime expires and reconciles overdue deployments. Treat the end time as a hard deadline for your work; do not rely on a grace period or notification delivery.

Copy datasets, checkpoints, notebooks, and outputs elsewhere before the end time. Instance storage is not a durable backup.

## Termination

Telegram, CLI, MCP, and API can request termination. It is destructive and forfeits unused paid runtime. The API acknowledges asynchronously; check status until termination completes.

Only terminate a deployment you intend to destroy. An agent should obtain your explicit confirmation first. A new rental does not restore the previous machine's files.

There is no documented self-service extension at launch. [Contact support](/essentials/support) if a deployment remains stuck, including its ID but no secrets.


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