tags to group requests by environment, workflow, or customer account identifier. Keep secrets and personal data out of labels.
Tag requests
For JSON endpoints, add:tags, such as production,site-import. A request accepts up to 20 tags, each 1–50 characters. ZDR requests omit tags from retained logs.
Inspect usage
The dashboard usage page filters charts and request logs together by key and tags. Monitors have persistent tags that label their runs and changes; request tags label a single API call.Check credits and usage history
Retrieve credit usage returns your balance, current-period usage, and next refill. It needslogs:read and costs no credits.
Retrieve usage history returns credits and requests over time, bucketed by hour, day, or month. History is kept for one year. It also needs
logs:read and costs no credits.
data holds one bucket per interval, oldest first, each with start, credits_used, and request_count; intervals without usage are included with zero totals. With group_by, each bucket adds groups, highest credits first, each with the value it covers, credits_used, and request_count. total_credits_used and total_request_count cover the whole range.
Read request logs
List request logs needslogs:read and returns entries newest first, including batch settlements and monitor runs.
Each entry has
request_id, timestamp (when the request completed), method, path, status_code, error_code (null on success), latency_ms (server-side processing time), credits_used, key_id, tags, and zdr (whether the request was made under zero data retention). The response’s page, limit, and has_more drive pagination.
Retrieve a request log returns the same entry plus input.query (query parameters as sent), input.body (the request body with credentials and uploaded content redacted), response (the retained JSON response with credentials redacted, or null when unavailable), and user_agent. An entry becomes retrievable a few seconds after its request completes, so a 404 NOT_FOUND for a request that just finished means ingestion has not caught up yet; retry after a few seconds.
Entries include HTTP requests and synthetic MONITOR and BATCH rows for runs and settlements. Their method is not always an HTTP verb, and their key_id can be null.
Logs are kept for up to one year. Retained input and response fields redact recognized credentials; redaction does not remove every kind of sensitive content. ZDR retains operational metadata only for web-content requests. Brand retrieval and Styleguide requests with ZDR enabled create no request-log entry.
Save request_id with application errors so you can match them to the API logs and support reports.