Limits

Requests per minute, per key

Each API key has its own budget (600 requests per minute by default). A key that exceeds it receives 429 RATE_LIMITED with a Retry-After header (seconds); other keys are not affected. Every response carries X-RateLimit-Limit and X-RateLimit-Remaining so you can slow down before you hit the wall.

The limit is applied per instance of the API, so the effective ceiling can be higher than the figure when several instances serve you. Treat it as a guide for your pacing, and back off on 429.

Daily limits per capability

The platform can cap a metric of a capability per day for your project (for example the messages you send). When a limit is reached the call answers 429 QUOTA_EXCEEDED with Retry-After set to the seconds until the next UTC midnight. A refused attempt does not use up your allowance.

Read your usage and your limits:

const usage = await trida.usage.get({ from: '2026-10-01', to: '2026-10-31' });
// usage.data: [{ day, capability, metric, count }], usage.limits: [{ capability, metric, dailyLimit }]

from and to are UTC days (YYYY-MM-DD), at most 92 days apart, 30 days by default. The metric <name>_refused counts attempts refused by a limit.

Providers

When a call needs a third party (a payment provider, a message gateway) and it fails, you get PROVIDER_TIMEOUT (10 seconds without an answer), PROVIDER_ERROR, or PROVIDER_UNAVAILABLE when the provider keeps failing and calls to it are suspended for a minute. Retry with the same Idempotency-Key: an operation is never repeated by the platform on its own unless it is safe to repeat.

Other limits

• 20 webhook endpoints per project.
• Webhook deliveries wait 10 seconds for your answer.
• An Idempotency-Key is up to 255 printable characters, remembered for 24 hours.