Answers

What are Heidi's API rate limits?

Per scope, per 60-second window: 600 reads, 120 asks, 120 tool calls, 60 writes. Hitting a limit returns 429 with Retry-After and X-RateLimit headers, never a silent drop.

Reads are generous on purpose: recall-heavy callers like coding agents stream lots of small queries. Writes are tight because each one mutates the graph.
Both official SDKs retry 429s automatically, honouring the server's Retry-After.

Related

What scopes do Heidi API tokens have?

Is there a Python SDK for Heidi?

Ask AI about Heidi:
Share this: