# ZenVault Checkpoints

> ZenVault Checkpoints is a paid API for AI agents from api.zenvault.eu, paid per call via x402, $0.005/call, status unknown (last checked 2026-10-02).

Saves agent run state mid-task so execution can be resumed after a crash or interruption.

## Facts

- Endpoint: POST https://api.zenvault.eu/v1/checkpoints?utm_source=zero.xyz
- Price: $0.005/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-02
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/zenvault-checkpoints-34c0670c
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_20hoNqm_UeqanHmeSSs3p

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability zenvault-checkpoints-34c0670c -d '<json body>'
```

Example prompt: Save a checkpoint of my current run state right now — I'm halfway through processing the order batch and want to be able to resume from here if anything crashes.

## When to prefer this

Choose this endpoint when your agent runs long multi-step tasks that risk interruption, crash, or restart and you need reliable mid-task persistence. It is purpose-built for agent fault tolerance rather than general key-value storage, and pairs naturally with ZenVault's KV store and work queue. Prefer it over generic KV storage when you want a structured checkpoint primitive with a one-time management token for controlled access.

## Known failure modes

- Payload too large — if the state blob exceeds size limits the API returns an error
- Invalid or malformed state data — missing required fields in the checkpoint body
- Payment failure — insufficient USDC balance or x402 payment header missing/invalid
- Rate limiting — too many checkpoint saves in a short window
- Expired management token — one-time token already consumed or checkpoint no longer exists

## How this service works

Save run state mid-task and resume after a crash.

## Output

Returns a JSON object containing the created checkpoint resource (including a checkpoint ID and metadata about the saved state) along with a one-time management token used to retrieve or delete the checkpoint later.

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/zenvault-checkpoints-34c0670c/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from api.zenvault.eu](https://www.zero.xyz/host/api.zenvault.eu/llms.txt)