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

# Idempotency and retries

> Retry accepted mutations without creating duplicate work.

Every scrape, export, or cancellation mutation needs an identifier of 16–128 printable non-space ASCII characters.

* REST: send it in `Idempotency-Key`.
* MCP: pass it as `client_request_id` in tool arguments.
* REST helper: set `client_request_id` in the request document.

Generate one UUID per intended mutation. Reuse the same identifier **and identical arguments** after a retryable transport error. Use a new identifier for a genuinely new operation. When switching REST and MCP for a retry, preserve the same value.

`idempotency_conflict` means an identifier was reused with different input. `idempotency_in_progress` means an earlier request is still being handled. Honor `Retry-After` when present; do not create another job to work around either response.

A timeout or connection loss does not prove the server rejected the request. Keep a durable mapping from your local task to its mutation identifier and returned job ID.