ByteScope

pm disable-user

changes the device

Switches a package off for one Android user without removing it — the reversible alternative to uninstalling.

Command
adb shell pm disable-user
Commands
Packages (pm)
Risk
changes the device

Syntax

pm disable-user [--user <id>] <package>
pm enable [--user <id>] <package>

Flags and arguments

TokenMeaning
--user <id>Which Android user to change. Worth passing explicitly, because the state is per user and so is the command that undoes it.

Examples

  • pm disable-user --user 0 com.manufacturer.bloat

    Switch a preinstalled app off for the primary user.

  • pm enable --user 0 com.manufacturer.bloat

    Switch it back on.

  • pm list packages -d --user 0

    Confirm what is disabled, from the platform rather than from memory.

Reading the output

Sample output

Package com.manufacturer.bloat new state: disabled-user

The command prints the state it read back, which is not always the state you asked for — and that is the whole reason to read this line rather than to check for the absence of the word "error". `pm enable` on a package the platform will not re-enable prints `new state: disabled-user` and exits zero. Compare the state to what you wanted.

Output fields

FieldMeaning
disabled-userOff because a user (or this command) switched it off.
disabledOff because the app or the framework switched it off.
disabled-until-usedOff until something needs it.
enabledExplicitly on.
defaultNo override: whatever the app's manifest declares, which is normally on.

What access it needs

An authorised `adb` connection; no root and no `su`. That is the whole reason this command is the standard answer for a preinstalled app you cannot delete — the enabled state is a per-user override the shell is allowed to set, so nothing is written to the read-only system partition and nothing has to be unlocked first. Two access details shape how the reply should be read. `--user <id>` names whose override is being changed, and a build unwilling to let the shell touch another user's state refuses rather than quietly editing your own. And the platform reserves the right to decline a re-enable: `pm enable` on a package it will not switch back on prints a state line and exits zero, so the exit status is not the signal here. The state named in the line is. `parseEnabledStateLine` also checks that the package in the answer is the package in the question, because a device that replied about something else replied to a different question.

Failure messages, decoded

MessageMeaning
Package com.example new state: disabled-userPrinted by `pm enable`, this is a refusal wearing a success line's clothes: the command ran, exited zero, and left the app switched off. `classifyPmEnabledChange` therefore compares the state that came back against the state that was asked for instead of scanning for the word error, which is the difference between a restore button that works and one that only claims to.
Warning: Unknown optionThe option loop did not recognise a flag — in practice `--user` on a build too old for it. `isFailureLine` counts this as a failure despite the milder word, because a run that did not understand your flag did not do what you asked, whatever it did instead.
Exception occurred while executing 'disable-user':The wrapper caught a `Throwable` and the following line names it. `IllegalArgumentException` and `NameNotFoundException` both arrive this way and both point at a misspelt or absent package; no state line is printed at all, which is how the classifier separates this from a state it simply did not want.

What it changes, and the way back

The app stops running, stops receiving broadcasts and disappears from the launcher for that user; nothing is deleted. `pm enable --user <id> <pkg>` switches it back on. Disabling a system service other apps depend on can break them in ways that only show up later, so change one package at a time and watch what stops working.

Version and vendor notes

  • `disable-user` sets the state that this command reports back, so a package somebody else had already disabled answers `disabled` or `disabled-until-used`. All three mean it is not running.
  • A success line naming a state other than the one you asked for is a failure with a calm voice. Reading the line is the difference between an honest restore button and a broken one.

We have a UI for this

Android Debloat Tool runs this command over USB and parses what comes back — a Chromium browser, no SDK, nothing uploaded.

Open Android Debloat Tool

Related commands

Commands that turn up in the same session as pm disable-user.

CommandWhat it does
pm uninstallRemoves a package, or removes it for one Android user only — which is how a preinstalled app is disposed of without root.
pm list packagesLists installed packages, optionally with the APK path, uid, version code and installer.
dumpsys packageEverything the package manager knows about one app: version, paths, flags, install source and per-user state.

More Packages (pm) commands

CommandWhat it does
pm pathPrints the absolute paths of the APK files that make up an installed package.
pm clearWipes an app's data — or, on Android 13 and later, only its cache.
pm trim-cachesAsks the system to drop cached files until the given amount of free space exists.
pm installInstalls one APK that is already on the device, which is what `adb install` does under the hood.
pm install-createOpens an install session so several APKs — a base plus its splits — are installed as one atomic unit.

Checked against AOSP, toybox and kernel sources while writing the parsers behind our Web ADB tools. Where a behaviour genuinely varies by Android version or manufacturer, this page says so instead of giving a number that would be wrong on half the phones out there.

Frequently asked questions

How do I run an adb shell command?

Switch on developer options and USB debugging on the phone, plug it in, then run adb shell <command> from a machine with platform-tools installed — or run adb shell on its own to get an interactive prompt and type the command without the prefix. If the command produces binary output, such as a screenshot, use adb exec-out instead of adb shell: the plain shell service attaches a terminal, which rewrites newline bytes and corrupts the data.

Do adb shell commands need root?

Not the ones on this reference. They run as the shell user, which can manage packages for its own Android user, read most dumpsys services, write the settings provider, inject input and read shared storage. Root is only needed for things this reference deliberately avoids — reading another app's private data (except through run-as on a debuggable build), writing system partitions, or reading /data/anr directly.

Why does the output look different on my phone?

Either the Android version or the manufacturer. Some flags and fields arrived in a specific release, and those are noted per command. The rest is OEM patching, which hits dumpsys hardest: a dump is debug output whose text is whatever the service author last printed, and manufacturers change it and backport their changes. Read fields by name rather than by position, and treat an unfamiliar shape as a difference rather than as a failure.

Can I run this from a browser instead of a terminal?

For many commands, yes — the Web ADB tools on this site talk to a phone over WebUSB from a Chromium-based desktop browser, with no SDK installed and nothing uploaded. Each page links to the tool that runs its command, when one exists. Two limits are worth knowing: only one program can hold the ADB interface at a time, so a desktop adb server has to be stopped first, and adb forward and adb reverse cannot work in a browser at all.

Would rather click than type? The Web ADB Toolkit runs commands like this from a browser over USB. Or browse every command.