Twitch · Rate Limits

Twitch Rate Limits

The Twitch Helix API protects services with a token-bucket algorithm. Each application receives separate buckets for app access tokens and user access tokens, and individual endpoints can charge non-default point values. User-token requests are scoped per client ID per user per minute. The Extensions API uses a separate rate-limit framework documented in the extensions guide.

Twitch Rate Limits is the machine-readable rate-limit profile for Twitch on the APIs.io network, conforming to the API Commons Rate Limits specification.

It captures 4 rate-limit definitions, measuring points_per_minute, points_per_request, and varies.

The profile also includes 3 backoff/retry policies defined and response codes documented for throttled.

Tagged areas include Streaming, Developer API, and Rate Limiting.

4 Limits Throttle: 429
StreamingDeveloper APIRate Limiting

Limits

App Access Token Bucket client-id (app token)
points_per_minute · minute
800
Default endpoint cost is 1 point per request; some endpoints charge more. A separate bucket per app access token applies.
User Access Token Bucket client-id per user (user token)
points_per_minute · minute
800
Limits apply per client ID per user per minute when using user OAuth tokens.
Per-Endpoint Variable Cost endpoint
points_per_request
see endpoint documentation
Some endpoints deduct more than 1 point per request; consult the specific endpoint reference for its point cost.
Extensions API extension/client
varies
see Extensions API documentation
The Extensions API has its own rate-limit framework documented separately.

Policies

Token Bucket
Each request deducts its endpoint's point value from the bucket; HTTP 429 is returned if the bucket would go below zero within the 1-minute window. The bucket replenishes at the documented rate.
Separate App vs User Buckets
App access token requests and user access token requests draw from independent buckets; you do not need to choose between them, but you must track each separately.
Header-Driven Recovery
Use Ratelimit-Remaining to back off proactively and Ratelimit-Reset (Unix timestamp) to wait the correct interval before retrying after a 429.

Sources