Review the usage plan for each SP-API operation in the official documentation; each operation defines a `rate` (requests per second added to a token bucket) and a `burst` (maximum token bucket size).
Read the `x-amzn-RateLimit-Limit` response header on successful (2xx, 400, 404) responses to get the current per-account-application rate limit for that operation; this header is absent on 401/403 responses.
When a 429 response is received, pause all requests to that operation and implement exponential backoff: start with a short delay (e.g., 1 second), double on each retry, and add random jitter to prevent synchronized retry storms.
Avoid bursting all available tokens at startup; throttle proactive polling loops to stay at or below the documented `rate` value to preserve burst capacity for urgent real-time operations.
For high-volume operations (e.g., catalog updates), prefer batch endpoints or Feeds API submissions over individual item calls to reduce per-unit API consumption.
Request a rate limit increase via the SP-API support portal for operations where documented rates are insufficient for your selling volume.
Known gotchas
Rate limits in SP-API are per account-application pair, not per seller account globally; a single seller using multiple integrations consumes separate token buckets per application.
The `x-amzn-RateLimit-Limit` header reflects the rate limit at the time of the response, which can vary based on account tier — do not hardcode limit values from documentation.
Retrying 429s without jitter causes synchronized retry waves (thundering herd) that can re-throttle the entire application; always add random jitter to backoff intervals.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp