Rate limits

Rate limiting applies to audit ingest and telemetry ingest routes when enabled in server configuration.

HTTP 429

When rate limited, the API returns 429 Too Many Requests with a Retry-After header (seconds). The SDK automatically retries on 429 and 5xx with exponential backoff (default 3 attempts).

SDK retry policy

Status	Behavior
429	Retry with backoff; respect `Retry-After`
500, 502, 503, 504	Retry with backoff
400, 401, 403, 409, 413	Fail immediately; repair input/configuration before retry

Batching defaults

Setting	Default
`batch_max_size`	100 events
`batch_max_wait_ms`	500 ms
Max batch on wire	1000 events

Batching reduces request volume and helps stay under rate limits.

Metrics

The platform exposes fact0_ratelimit_dropped_total on /metrics for observability.

Request and capture limits

The core limits encoded request bodies to 4 MiB. HTTP 200 may include per-item errors; inspect accepted/rejected counts. The Claude collector splits batches and retains permanent oversized failures locally within its configured spool limit. The core uses synchronous ingestion; receipt polling is unavailable.