Rate limits

The default

300 requests per minute. Crossing it returns HTTP 429.

Tighter limits

Some routes cost more to serve or touch the network, and carry their own limit. All are per minute.

Limit	Kind of route
120	Batch stamp detail lookups
30	Batch token price lookups
20	Every challenge request, deep health, ecosystem health, and batch dispenser lookups
10	Fee bump by child pays for parent, and atomic offer preparation
6	Alert scans
5	Broadcasting a transaction, and rebroadcasting one

The pattern is consistent: reads are generous, batch reads are moderate, anything that signs, broadcasts, or issues a challenge is deliberately slow.

Exempt routes

Media routes do not count against any limit:

• raw stamp bytes
• stamp previews and thumbnails
• token logos

They are cacheable and content addressed. See Media endpoints.

Staying under the limit

Cache on your side. Server-side caches already hold most responses for 20 seconds to a few minutes, so polling faster than that returns the same bytes and spends your budget for nothing. See Freshness.

Batch where a batch route exists. Stamp details take up to 24 ids in one call, and token prices take a list of tickers. Two batch calls beat forty single ones.

Use conditional requests on media. Previews honour If-None-Match and answer 304.

Ask for what you need. The token list accepts limit up to 500. Pulling everything and filtering client side is the expensive way round.

When you get a 429

Wait a minute and slow down. The limit is per minute, not a rolling penalty, so a pause clears it. Retrying immediately in a loop just keeps you at the limit.

Related

• API quick start
• Worked examples