RTC rate limits
The RTC API limits how many requests it accepts in a time window. Limits protect every customer’s rooms from a single runaway client; they are not a billing meter.
What is limited
Section titled “What is limited”| Limit | Counts | Purpose |
|---|---|---|
| Per IP address | Every request from one address, signed in or not | Abuse protection; the lowest layer |
| Per API key | Requests made with one project API key | Keeps one integration or leaked key from starving the others |
| Per organization | Requests of all keys and console sessions of one RTC organization | Fair share across customers |
A request must pass every limit that applies to it. The per-key and per-organization limits depend on your plan, and CHAMPREP can adjust them for an organization. Read the headers below instead of hard-coding numbers.
Some operations have their own, narrower limits that answer with a specific
code, for example webhook replays (REPLAY_QUOTA)
and in-room app messages (MESSAGE_RATE_LIMITED).
Headers
Section titled “Headers”Authenticated responses carry the state of the tightest limit that applies:
| Header | Meaning |
|---|---|
X-RateLimit-Limit | Requests allowed in the current window |
X-RateLimit-Remaining | Requests left in the current window |
X-RateLimit-Reset | Seconds until the window resets |
A request over a limit answers 429 with the code
RATE_LIMITED and a Retry-After header, in
seconds.
HTTP/1.1 429 Too Many RequestsRetry-After: 12X-RateLimit-Limit: 600X-RateLimit-Remaining: 0X-RateLimit-Reset: 12X-Error-Code: RATE_LIMITEDX-Request-Id: 6f1c2b8e-1d2a-4c55-9a8e-0f4b9b1d7c21The numbers above are illustrative.
Back off correctly
Section titled “Back off correctly”- On
429, wait at leastRetry-Afterseconds before the next request, then resume gradually. Do not retry in a tight loop. - Spread periodic work (polling, reconciliation) over time, and prefer webhooks to polling.
- Use one API key per integration: its own limit then contains a runaway job, and you can revoke one key without affecting the others.
- Mint participant credentials when a user actually joins, not in advance. For high join volumes, mint join tokens offline with a project signing key.
- Use
?limitand?cursorto page through lists rather than repeating full list requests.