Realtime: ClientPresenceRateLimitReached error

If a client sends Presence updates too frequently, you may see this error code in your Realtime logs:

1
ClientPresenceRateLimitReached

On the client side, the channel receives a system error message and is then closed. Any Presence, Broadcast, or Postgres Changes subscriptions on that channel stop until the client reconnects.

This error almost always means Presence is being used for high-frequency updates that it isn't designed for. This guide explains the limit, why it exists, and how to fix it.

Why the error occurs#

Each client has a per-connection limit on how often it can send Presence updates. By default, a client can send at most 5 Presence updates within a 30-second window. Both track() and untrack() calls count toward this limit. When a client exceeds it, Realtime logs ClientPresenceRateLimitReached and shuts the channel down.

This is a per-client safeguard, and it is separate from the project-wide presence events per second limit (logged as PresenceRateLimitReached). A single client can hit ClientPresenceRateLimitReached on its own, even when overall project usage is low.

The limit exists because Presence syncs state through the server and notifies every subscriber on the channel on each change. A client that calls track() in a tight loop—for example, on every mouse move to share a cursor position—multiplies its updates across all subscribers and degrades the channel for everyone. The rate limit stops one client from doing this.

How to fix it#

The fix is to stop sending frequent Presence updates. Choose the option that matches your use case.

Use Broadcast for high-frequency updates#

Presence is meant for slow-changing state such as online/offline status, the document a user is viewing, or which page they're on. For high-frequency or fire-and-forget data—live cursors, typing indicators, pointer positions—use Broadcast instead. Broadcast relays messages through Realtime to connected clients without maintaining synced Presence state, so it handles rapid updates without triggering this limit.

Throttle your Presence updates#

If you do need Presence for state that changes often, throttle your track() and untrack() calls so the client sends at most a few Presence updates per window. Only update Presence when the shared state changes, and coalesce bursts into a single update rather than sending one on every event.

How to prevent it#

• Reserve Presence for slow-changing state, and reach for Broadcast for anything that updates rapidly. See the guidance in the Presence guide.
• Call track() only when the shared state changes, not on a timer or on every input event.

Related resources#

• Presence — when to use Presence and how it works
• Broadcast — the right tool for high-frequency updates
• Realtime limits — other limits that apply to your project