Fix guide · updated 2026-11-16
Supabase "Request rate limit reached" (429)
AuthApiError: Request rate limit reached · over_request_rate_limit · 429
Sign-ins start failing with "Request rate limit reached", HTTP 429. Usually while you test, sometimes for all your users at once. Supabase Auth limits how often one IP address can call each of its endpoints. We ran Supabase's real Auth server with its default limits and supabase-js 2.119 on our machine to see which calls count, how fast they come back, and what a 429 does to someone who is already signed in.
What the error means
Supabase Auth gives each IP address an allowance per endpoint. It holds about 30 requests and refills over time. When it's empty, the call gets HTTP 429 with the code over_request_rate_limit:
429 {"code":"over_request_rate_limit","message":"Request rate limit reached"}
In our test, the answer had no Retry-After header to say when to try again. What we measured with the default limits:
- Password sign-ins and session refreshes (
/token) share one allowance: about 30 at once, then one more every 2 seconds (150 per 5 minutes). Wrong passwords count too: after 30 wrong ones, the right password got the 429 as well. - Sign-ups (
/signup): 30 at once, then one every 10 seconds (30 per 5 minutes). updateUser(/user) and email codes and links (verifyOtp,/verify): 30 at once each.- Anonymous sign-ins: 30 per hour, see our anonymous sign-ins guide.
Each endpoint counts on its own: once sign-ups ran out, a password reset, a magic link, updateUser and a password sign-in from the same address still went through. Each IP address counts on its own too: the same sign-in from another address worked at once. The numbers are under Authentication → Rate Limits in your Supabase dashboard, where most of them can be raised.
1. You're testing sign-in or sign-up over and over
Thirty quick tries from your machine and that endpoint is closed. Automated tests get there first. You don't need to wait the full 5 minutes for one more try: a sign-in comes back every 2 seconds, a sign-up every 10. For a test suite, raise the limits under Authentication → Rate Limits, or space the calls out.
2. Your server makes the calls for everyone
If sign-in, sign-up or the session refresh runs on your server (a Next.js server action, an API route, middleware, an edge function), Supabase sees your server's IP address, not your visitors'. Then all your users share one allowance. In our test, 30 quick sign-ins from one address closed sign-in for everyone else at that address.
Supabase's fix is IP address forwarding. Turn on IP Address Forwarding under Authentication → Rate Limits, then have your server send each visitor's IP address in an Sb-Forwarded-For header on its Auth calls, as Supabase's rate-limit docs show:
global: { headers: { 'sb-forwarded-for': request.headers.get('x-forwarded-for') } }
Supabase's docs say the header only counts on requests made with a secret key (sb_secret_…), not the publishable key or the legacy anon and service_role keys. Keep that key in server code only. In our test, with forwarding on, Auth counted each forwarded address separately: 60 sign-ins from one server, each forwarding a different visitor's address, all went through.
3. Many people share one address
An office, a school or a mobile carrier can put many people behind one IP address, and they share every allowance above. Session refreshes count in the same allowance as password sign-ins, so a page that refreshes the session far more often than it needs to uses it up for everyone on that network. Open your browser's Network tab, filter on /auth/v1/, and count the token requests per page load: more than one is a loop to fix.
4. A 429 can sign people out
This is the costly case. In our test, a supabase-js 2.119 client whose access token had expired asked for a new one while the allowance was empty. getSession() came back with no session and the 429 error, and the client fired SIGNED_OUT. The person has to sign in again, which needs the same allowance. With some allowance left, the same steps gave TOKEN_REFRESHED.
Reading supabase-js's code, it only drops the session when the access token has already expired. A refresh that fails while the token is still valid keeps the session and tries again later. So leave autoRefreshToken on (the default in the browser), and the refresh happens before the token runs out.
If an AI tool built your app
Give your AI tool this prompt:
In every Supabase Auth call (signInWithPassword, signUp, verifyOtp, updateUser, resetPasswordForEmail), when the error code is over_request_rate_limit (HTTP 429), show "Too many attempts. Please wait a minute and try again." instead of the raw error, and disable the submit button for 30 seconds. Make sure the app creates one Supabase client and reuses it, and that nothing calls signIn, signUp, refreshSession or updateUser in a loop or on every render. Don't change anything else.
If your sign-in runs on a server, the forwarding setting and the secret key (case 2) are yours to set up: the AI can't open your dashboard.
Not quite your error?
- "email rate limit exceeded", or "For security purposes, you can only request this after … seconds": Supabase "email rate limit exceeded".
- Users are signed out with "Invalid Refresh Token": Supabase "Invalid Refresh Token: Refresh Token Not Found".
- Guest users can't sign in at all: Supabase "Anonymous sign-ins are disabled".
- Sign-in fails with the right password and no 429: Supabase "Invalid login credentials".
Still stuck?
Find out which allowance runs out. In your browser's Network tab, the refused request's path says it: token (sign-ins and refreshes), signup, user or verify. If it happens to your users and not to you, and your sign-in runs on a server, start with case 2.
Other Supabase problems? See Supabase fixes.