cloudd — the daemon that syncs app data through iCloud

Last updated: 2026-07-30

What it is

cloudd is the daemon behind CloudKit. Where iCloud Drive deals in files, CloudKit deals in structured records — the kind of data an app holds internally and wants mirrored across your devices.

A surprising amount runs on it. Apple's own Notes, Reminders, Freeform and the iCloud Keychain's record layer are CloudKit clients. So are a very large number of third-party apps: when a developer wants sync without running their own server or asking you to create an account, CloudKit is the obvious choice, because the storage is billed to your iCloud allowance and the identity is the Apple Account you already have.

The consequence is that cloudd's traffic is an aggregate. It is not one feature's sync; it is every CloudKit-based app on your Mac sharing one pipe.

Why it talks to the network

Three distinct triggers, and telling them apart explains most confusing readings.

You changed something. A note edited, a reminder ticked, an app's data modified. The change is pushed up. These are small and frequent.

Something changed elsewhere. An edit on your iPhone arrives on your Mac through cloudd. This is why the process can be busy on a machine you have not touched — the work originated on another device.

A dataset is being established. Enabling iCloud for an app that already holds a lot of data means the entire existing set has to be uploaded once. Signing in to a new Mac means the reverse: the entire set comes down. Both are one-time bursts that can be large, and both are frequently mistaken for a runaway process.

How much traffic is normal

Day to day, modest — typically single-digit megabytes for ordinary editing, because CloudKit syncs records rather than whole files and the records are small.

The exceptions are the establishment cases above. Turning on sync for an app with years of accumulated data, or setting up a new Mac, can produce hundreds of megabytes to several gigabytes in a single session. That is expected behaviour and it should subside once the initial reconciliation completes.

The pattern to be suspicious of is sustained volume with no corresponding change — cloudd busy for days at a stretch on a Mac where nothing much is happening. That usually indicates a sync that is failing and retrying, and the fix is generally at the app or account level rather than the network level.

Can you turn it off

Not directly, and there is a better lever. cloudd is a throttleOK process: blocking it is disruptive, but limiting it is safe.

There is no supported switch that disables the daemon, and killing it accomplishes nothing durable. What you can control is what it has to carry. System Settings › your Apple Account › iCloud lists the apps syncing through iCloud, and turning off the ones you do not care about removes their traffic at the source. This is more effective than any network-level intervention, because the data is never queued in the first place.

When the issue is timing rather than the sync itself — you need the bandwidth right now, not permanently — throttling is the correct tool. CloudKit is built to tolerate slow and interrupted connections, so a rate limit delays the sync without breaking it. The records arrive later; nothing is lost.

Seeing what it actually used

cloudd's volume only makes sense against a timeline. A one-off multi-gigabyte upload the day you enabled a sync is completely normal; the same figure appearing every day is not. Without history, those two look identical.

Bytetally keeps per-process upload and download over time, so you can see whether cloudd's spike was an event or a pattern — and if it is a pattern, cap its rate while leaving the sync alive rather than cutting it off and wondering later why something never synced.

Related processes

Common questions