General protocol syntax

Protocol overview

The MPD command protocol exchanges line-based text records between client and server over TCP. Once the client is connected to the server, they conduct a conversation until the client closes the connection. The conversation flow is always initiated by the client.

All data between the client and the server is encoded in UTF-8.

The client transmits a command sequence, terminated by the newline character \n. The server will respond with one or more lines, the last of which will be a completion code.

When the client connects to the server, the server will answer with the following line:

OK MPD version

where version is a version identifier such as 0.12.2. This version identifier is the version of the protocol spoken, not the real version of the daemon. (There is no way to retrieve this real version identifier from the connection.)



If arguments contain spaces, they should be surrounded by double quotation marks.

Argument strings are separated from the command and any other arguments by linear white-space (’ ‘ or ‘\t’).


A command returns OK on completion or ACK some error on failure. These denote the end of command execution.

Some commands return more data before the response ends with OK. Each line is usually in the form NAME: VALUE. Example:

foo: bar

Binary Responses

Some commands can return binary data. This is initiated by a line containing binary: 1234 (followed as usual by a newline). After that, the specified number of bytes of binary data follows, then a newline, and finally the OK line.

If the object to be transmitted is large, the server may choose a reasonable chunk size and transmit only a portion. The maximum chunk size can be changed by clients with the binarylimit command.

Usually, the response also contains a size line which specifies the total (uncropped) size, and the command usually has a way to specify an offset into the object; this way, the client can copy the whole file without blocking the connection for too long.


foo: bar
binary: 42
<42 bytes>

Failure responses

The nature of the error can be gleaned from the information that follows the ACK. ACK lines are of the form:

ACK [error@command_listNum] {current_command} message_text

These responses are generated by a call to commandError. They contain four separate terms. Let’s look at each of them:

  • error: numeric value of one of the ACK_ERROR constants defined in src/protocol/Ack.hxx.

  • command_listNum: offset of the command that caused the error in a Command List. An error will always cause a command list to terminate at the command that causes the error.

  • current_command: name of the command, in a Command List, that was executing when the error occurred.

  • message_text: some (hopefully) informative text that describes the nature of the error.

An example might help. Consider the following sequence sent from the client to the server:

volume 86
play 10240

The server responds with:

ACK [50@1] {play} song doesn't exist: "10240"

This tells us that the play command, which was the second in the list (the first or only command is numbered 0), failed with error 50. The number 50 translates to ACK_ERROR_NO_EXIST – the song doesn’t exist. This is reiterated by the message text which also tells us which song doesn’t exist.

Command lists

To facilitate faster adding of files etc. you can pass a list of commands all at once using a command list. The command list begins with command_list_begin or command_list_ok_begin and ends with command_list_end.

It does not execute any commands until the list has ended. The response is a concatentation of all individual responses. On success for all commands, OK is returned. If a command fails, no more commands are executed and the appropriate ACK error is returned. If command_list_ok_begin is used, list_OK is returned for each successful command executed in the command list.


Some commands (e.g. delete) allow specifying a range in the form START:END (the END item is not included in the range, similar to ranges in the Python programming language). If END is omitted, then the maximum possible value is assumed.


All commands which search for songs (e.g. find and searchadd) share a common filter syntax:


EXPRESSION is a string enclosed in parentheses which can be one of:

  • (TAG == 'VALUE'): match a tag value; if there are multiple values of the given type, at least one must match. (TAG != 'VALUE'): mismatch a tag value; if there are multiple values of the given type, none of them must match. The special tag any checks all tag types. AlbumArtist looks for VALUE in AlbumArtist and falls back to Artist tags if AlbumArtist does not exist. VALUE is what to find. An empty value string means: match only if the given tag type does not exist at all; this implies that negation with an empty value checks for the existence of the given tag type.

  • (TAG contains 'VALUE') checks if the given value is a substring of the tag value.

  • (TAG =~ 'VALUE') and (TAG !~ 'VALUE') use a Perl-compatible regular expression instead of doing a simple string comparison. (This feature is only available if MPD was compiled with libpcre)

  • (file == 'VALUE'): match the full song URI (relative to the music directory).

  • (base 'VALUE'): restrict the search to songs in the given directory (relative to the music directory).

  • (modified-since 'VALUE'): compares the file’s time stamp with the given value (ISO 8601 or UNIX time stamp).

  • (AudioFormat == 'SAMPLERATE:BITS:CHANNELS'): compares the audio format with the given value. See Global Audio Format for a detailed explanation.

  • (AudioFormat =~ 'SAMPLERATE:BITS:CHANNELS'): matches the audio format with the given mask (i.e. one or more attributes may be *).

  • (!EXPRESSION): negate an expression. Note that each expression must be enclosed in parentheses, e.g. (!(artist == 'VALUE')) (which is equivalent to (artist != 'VALUE'))

  • (EXPRESSION1 AND EXPRESSION2 ...): combine two or more expressions with logical “and”. Note that each expression must be enclosed in parentheses, e.g. ((artist == 'FOO') AND (album == 'BAR'))

The find commands are case sensitive, while search and related commands ignore case.

Prior to MPD 0.21, the syntax looked like this:


Escaping String Values

String values are quoted with single or double quotes, and special characters within those values must be escaped with the backslash (\). Keep in mind that the backslash is also the escape character on the protocol level, which means you may need to use double backslash.

Example expression which matches an artist named foo'bar":

(Artist == "foo\'bar\"")

At the protocol level, the command must look like this:

find "(Artist == \"foo\\'bar\\\"\")"

The double quotes enclosing the artist name must be escaped because they are inside a double-quoted find parameter. The single quote inside that artist name must be escaped with two backslashes; one to escape the single quote, and another one because the backslash inside the string inside the parameter needs to be escaped as well. The double quote has three confusing backslashes: two to build one backslash, and another one to escape the double quote on the protocol level. Phew!

To reduce confusion, you should use a library such as libmpdclient which escapes command arguments for you.


The following tags are supported by MPD:

  • artist: the artist name. Its meaning is not well-defined; see “composer” and “performer” for more specific tags.

  • artistsort: same as artist, but for sorting. This usually omits prefixes such as “The”.

  • album: the album name.

  • albumsort: same as album, but for sorting.

  • albumartist: on multi-artist albums, this is the artist name which shall be used for the whole album. The exact meaning of this tag is not well-defined.

  • albumartistsort: same as albumartist, but for sorting.

  • title: the song title.

  • track: the decimal track number within the album.

  • name: a name for this song. This is not the song title. The exact meaning of this tag is not well-defined. It is often used by badly configured internet radio stations with broken tags to squeeze both the artist name and the song title in one tag.

  • genre: the music genre.

  • date: the song’s release date. This is usually a 4-digit year.

  • originaldate: the song’s original release date.

  • composer: the artist who composed the song.

  • composersort: same as composer, but for sorting.

  • performer: the artist who performed the song.

  • conductor: the conductor who conducted the song.

  • work: “a work is a distinct intellectual or artistic creation, which can be expressed in the form of one or more audio recordings”

  • ensemble: the ensemble performing this song, e.g. “Wiener Philharmoniker”.

  • movement: name of the movement, e.g. “Andante con moto”.

  • movementnumber: movement number, e.g. “2” or “II”.

  • location: location of the recording, e.g. “Royal Albert Hall”.

  • grouping: “used if the sound belongs to a larger category of sounds/music” (from the IDv2.4.0 TIT1 description).

  • comment: a human-readable comment about this song. The exact meaning of this tag is not well-defined.

  • disc: the decimal disc number in a multi-disc album.

  • label: the name of the label or publisher.

  • musicbrainz_artistid: the artist id in the MusicBrainz database.

  • musicbrainz_albumid: the album id in the MusicBrainz database.

  • musicbrainz_albumartistid: the album artist id in the MusicBrainz database.

  • musicbrainz_trackid: the track id in the MusicBrainz database.

  • musicbrainz_releasetrackid: the release track id in the MusicBrainz database.

  • musicbrainz_workid: the work id in the MusicBrainz database.

There can be multiple values for some of these tags. For example, MPD may return multiple lines with a performer tag. A tag value is a UTF-8 string.

Other Metadata

The response to lsinfo and similar commands may contain song tags and other metadata, specifically:

  • duration: the duration of the song in seconds; may contain a fractional part.

  • time: like duration, but as integer value. This is deprecated and is only here for compatibility with older clients. Do not use.

  • Range: if this is a queue item referring only to a portion of the song file, then this attribute contains the time range in the form START-END or START- (open ended); both START and END are time stamps within the song in seconds (may contain a fractional part). Example: 60-120 plays only the second minute; “180 skips the first three minutes.

  • Format: the audio format of the song (or an approximation to a format supported by MPD and the decoder plugin being used). When playing this file, the audio value in the status response should be the same.

  • Last-Modified: the time stamp of the last modification of the underlying file in ISO 8601 format. Example: “2008-09-28T20:04:57Z



Often, users run MPD with random enabled, but want to be able to insert songs “before” the rest of the playlist. That is commonly called “queuing”.

MPD implements this by allowing the client to specify a “priority” for each song in the playlist (commands priod and priodid). A higher priority means that the song is going to be played before the other songs.

In “random” mode, MPD maintains an internal randomized sequence of songs. In this sequence, songs with a higher priority come first, and all songs with the same priority are shuffled (by default, all songs are shuffled, because all have the same priority “0”). When you increase the priority of a song, it is moved to the front of the sequence according to its new priority, but always after the current one. A song that has been played already (it’s “before” the current song in that sequence) will only be scheduled for repeated playback if its priority has become bigger than the priority of the current song. Decreasing the priority of a song will move it farther to the end of the sequence. Changing the priority of the current song has no effect on the sequence. During playback, a song’s priority is reset to zero.

Command reference


For manipulating playlists and playing, there are two sets of commands. One set uses the song id of a song in the playlist, while another set uses the playlist position of the song. The commands using song ids should be used instead of the commands that manipulate and control playback based on playlist position. Using song ids is a safer method when multiple clients are interacting with MPD.

Querying MPD’s status


Clears the current error message in status (this is also accomplished by any command that starts playback).


Displays the song info of the current song (same song that is identified in status). Information about the current song is represented by key-value pairs, one on each line. The first pair must be the file key-value pair.

idle [SUBSYSTEMS...] [1]

Waits until there is a noteworthy change in one or more of MPD’s subsystems. As soon as there is one, it lists all changed systems in a line in the format changed: SUBSYSTEM, where SUBSYSTEM is one of the following:

  • database: the song database has been modified after update.

  • update: a database update has started or finished. If the database was modified during the update, the database event is also emitted.

  • stored_playlist: a stored playlist has been modified, renamed, created or deleted

  • playlist: the queue (i.e. the current playlist) has been modified

  • player: the player has been started, stopped or seeked or tags of the currently playing song have changed (e.g. received from stream)

  • mixer: the volume has been changed

  • output: an audio output has been added, removed or modified (e.g. renamed, enabled or disabled)

  • options: options like repeat, random, crossfade, replay gain

  • partition: a partition was added, removed or changed

  • sticker: the sticker database has been modified.

  • subscription: a client has subscribed or unsubscribed to a channel

  • message: a message was received on a channel this client is subscribed to; this event is only emitted when the queue is empty

  • neighbor: a neighbor was found or lost

  • mount: the mount list has changed

Change events accumulate, even while the connection is not in “idle” mode; no events get lost while the client is doing something else with the connection. If an event had already occurred since the last call, the new idle command will return immediately.

While a client is waiting for idle results, the server disables timeouts, allowing a client to wait for events as long as mpd runs. The idle command can be canceled by sending the command noidle (no other commands are allowed). MPD will then leave idle mode and print results immediately; might be empty at this time. If the optional SUBSYSTEMS argument is used, MPD will only send notifications when something changed in one of the specified subsytems.


Reports the current status of the player and the volume level.

  • partition: the name of the current partition (see Partition commands)

  • volume: 0-100 (deprecated: -1 if the volume cannot be determined)

  • repeat: 0 or 1

  • random: 0 or 1

  • single [2]: 0, 1, or oneshot [6]

  • consume [2]: 0 or 1

  • playlist: 31-bit unsigned integer, the playlist version number

  • playlistlength: integer, the length of the playlist

  • state: play, stop, or pause

  • song: playlist song number of the current song stopped on or playing

  • songid: playlist songid of the current song stopped on or playing

  • nextsong [2]: playlist song number of the next song to be played

  • nextsongid [2]: playlist songid of the next song to be played

  • time: total time elapsed (of current playing/paused song) in seconds (deprecated, use elapsed instead)

  • elapsed [3]: Total time elapsed within the current song in seconds, but with higher resolution.

  • duration [5]: Duration of the current song in seconds.

  • bitrate: instantaneous bitrate in kbps

  • xfade: crossfade in seconds (see Cross-Fading)

  • mixrampdb: mixramp threshold in dB

  • mixrampdelay: mixrampdelay in seconds

  • audio: The format emitted by the decoder plugin during playback, format: samplerate:bits:channels. See Global Audio Format for a detailed explanation.

  • updating_db: job id

  • error: if there is an error, returns message here

MPD may omit lines which have no (known) value. Older MPD versions used to have a “magic” value for “unknown”, e.g. “volume: -1”.


Displays statistics.

  • artists: number of artists

  • albums: number of albums

  • songs: number of songs

  • uptime: daemon uptime in seconds

  • db_playtime: sum of all song times in the database in seconds

  • db_update: last db update in UNIX time (seconds since 1970-01-01 UTC)

  • playtime: time length of music played

Playback options

consume {STATE} [2]

Sets consume state to STATE, STATE should be 0 or 1. When consume is activated, each song played is removed from playlist.

crossfade {SECONDS}

Sets crossfading between songs. See Cross-Fading.

mixrampdb {deciBels}

Sets the threshold at which songs will be overlapped. See MixRamp.

mixrampdelay {SECONDS}

Additional time subtracted from the overlap calculated by mixrampdb. A value of “nan” disables MixRamp overlapping and falls back to crossfading. See MixRamp.

random {STATE}

Sets random state to STATE, STATE should be 0 or 1.

repeat {STATE}

Sets repeat state to STATE, STATE should be 0 or 1.

If enabled, MPD keeps repeating the whole queue (single mode disabled) or the current song (single mode enabled).

If random mode is also enabled, the playback order will be shuffled each time the queue gets repeated.

setvol {VOL}

Sets volume to VOL, the range of volume is 0-100.

getvol [8]

Read the volume. The result is a volume: line like in status. If there is no mixer, MPD will emit an empty response. Example:

volume: 42
single {STATE} [2]

Sets single state to STATE, STATE should be 0, 1 or oneshot [6]. When single is activated, playback is stopped after current song, or song is repeated if the ‘repeat’ mode is enabled.

replay_gain_mode {MODE} [3]

Sets the replay gain mode. One of off, track, album, auto . Changing the mode during playback may take several seconds, because the new settings do not affect the buffered data. This command triggers the options idle event.


Prints replay gain options. Currently, only the variable replay_gain_mode is returned.

volume {CHANGE}

Changes volume by amount CHANGE. Deprecated, use setvol instead.

Controlling playback


Plays next song in the playlist.

pause {STATE}

Pause or resume playback. Pass 1 to pause playback or 0 to resume playback. Without the parameter, the pause state is toggled.

play [SONGPOS]

Begins playing the playlist at song number SONGPOS.

playid [SONGID]

Begins playing the playlist at song SONGID.


Plays previous song in the playlist.


Seeks to the position TIME (in seconds; fractions allowed) of entry SONGPOS in the playlist.

seekid {SONGID} {TIME}

Seeks to the position TIME (in seconds; fractions allowed) of song SONGID.

seekcur {TIME}

Seeks to the position TIME (in seconds; fractions allowed) within the current song. If prefixed by + or -, then the time is relative to the current playing position.


Stops playing.

The Queue


The “queue” used to be called “current playlist” or just “playlist”, but that was deemed confusing, because “playlists” are also files containing a sequence of songs. Those “playlist files” or “stored playlists” can be loaded into the queue and the queue can be saved into a playlist file, but they are not to be confused with the queue.

Many of the command names in this section reflect the old naming convention, but for the sake of compatibility, we cannot rename commands.

There are two ways to address songs within the queue: by their position and by their id.

The position is a 0-based index. It is unstable by design: if you move, delete or insert songs, all following indices will change, and a client can never be sure what song is behind a given index/position.

Song ids on the other hand are stable: an id is assigned to a song when it is added, and will stay the same, no matter how much it is moved around. Adding the same song twice will assign different ids to them, and a deleted-and-readded song will have a new id. This way, a client can always be sure the correct song is being used.

Many commands come in two flavors, one for each address type. Whenever possible, ids should be used.


Adds the file URI to the playlist (directories add recursively). URI can also be a single file.

The position parameter is the same as in addid. [10]

Clients that are connected via local socket may add arbitrary local files (URI is an absolute path). Example:

add "/home/foo/Music/bar.ogg"
addid {URI} [POSITION]

Adds a song to the playlist (non-recursive) and returns the song id. URI is always a single file or URL. For example:

addid "foo.mp3"
Id: 999

If the second parameter is given, then the song is inserted at the specified position. If the parameter starts with + or -, then it is relative to the current song [8]; e.g. +0 inserts right after the current song and -0 inserts right before the current song (i.e. zero songs between the current song and the newly added song).


Clears the queue.

delete [{POS} | {START:END}]

Deletes a song from the playlist.

deleteid {SONGID}

Deletes the song SONGID from the playlist

move [{FROM} | {START:END}] {TO}

Moves the song at FROM or range of songs at START:END [2] to TO in the playlist.

If TO starts with + or -, then it is relative to the current song; e.g. +0 moves to right after the current song and -0 moves to right before the current song (i.e. zero songs between the current song and the moved range).

moveid {FROM} {TO}

Moves the song with FROM (songid) to TO (playlist index) in the playlist.

If TO starts with + or -, then it is relative to the current song; e.g. +0 moves to right after the current song and -0 moves to right before the current song (i.e. zero songs between the current song and the moved song).


Displays the queue.

Do not use this, instead use playlistinfo.

playlistfind {FILTER}

Search the queue for songs matching FILTER (see Filters).

playlistid {SONGID}

Displays a list of songs in the playlist. SONGID is optional and specifies a single song to display info for.

playlistinfo [[SONGPOS] | [START:END]]

Displays a list of all songs in the playlist, or if the optional argument is given, displays information only for the song SONGPOS or the range of songs START:END [2]

playlistsearch {FILTER}

Search the queue for songs matching FILTER (see Filters). Parameters have the same meaning as for find, except that search is not case sensitive.

plchanges {VERSION} [START:END]

Displays changed songs currently in the playlist since VERSION. Start and end positions may be given to limit the output to changes in the given range.

To detect songs that were deleted at the end of the playlist, use playlistlength returned by status command.

plchangesposid {VERSION} [START:END]

Displays changed songs currently in the playlist since VERSION. This function only returns the position and the id of the changed song, not the complete metadata. This is more bandwidth efficient.

To detect songs that were deleted at the end of the playlist, use playlistlength returned by status command.


Set the priority of the specified songs. A higher priority means that it will be played first when “random” mode is enabled.

A priority is an integer between 0 and 255. The default priority of new songs is 0.

prioid {PRIORITY} {ID...}

Same as priod, but address the songs with their id.

rangeid {ID} {START:END} [4]

Since MPD 0.19 Specifies the portion of the song that shall be played. START and END are offsets in seconds (fractional seconds allowed); both are optional. Omitting both (i.e. sending just “:”) means “remove the range, play everything”. A song that is currently playing cannot be manipulated this way.

shuffle [START:END]

Shuffles the queue. START:END is optional and specifies a range of songs.

swap {SONG1} {SONG2}

Swaps the positions of SONG1 and SONG2.

swapid {SONG1} {SONG2}

Swaps the positions of SONG1 and SONG2 (both song ids).

addtagid {SONGID} {TAG} {VALUE}

Adds a tag to the specified song. Editing song tags is only possible for remote songs. This change is volatile: it may be overwritten by tags received from the server, and the data is gone when the song gets removed from the queue.

cleartagid {SONGID} [TAG]

Removes tags from the specified song. If TAG is not specified, then all tag values will be removed. Editing song tags is only possible for remote songs.

Stored playlists

Playlists are stored inside the configured playlist directory. They are addressed with their file name (without the directory and without the .m3u suffix).

Some of the commands described in this section can be used to run playlist plugins instead of the hard-coded simple m3u parser. They can access playlists in the music directory (relative path including the suffix), playlists in arbitrary location (absolute path including the suffix; allowed only for clients that are connected via local socket), or remote playlists (absolute URI with a supported scheme).

listplaylist {NAME}

Lists the songs in the playlist. Playlist plugins are supported.

listplaylistinfo {NAME}

Lists the songs with metadata in the playlist. Playlist plugins are supported.


Prints a list of the playlist directory. After each playlist name the server sends its last modification time as attribute “Last-Modified” in ISO 8601 format. To avoid problems due to clock differences between clients and the server, clients should not compare this value with their local clock.


Loads the playlist into the current queue. Playlist plugins are supported. A range may be specified to load only a part of the playlist.

The POSITION parameter specifies where the songs will be inserted into the queue; it can be relative as described in addid. (This requires specifying the range as well; the special value 0: can be used if the whole playlist shall be loaded at a certain queue position.) [9]

playlistadd {NAME} {URI} [POSITION]

Adds URI to the playlist NAME.m3u. NAME.m3u will be created if it does not exist.

The POSITION parameter specifies where the songs will be inserted into the playlist. [10]

playlistclear {NAME}

Clears the playlist NAME.m3u.

playlistdelete {NAME} {SONGPOS}

Deletes SONGPOS from the playlist NAME.m3u.

The second parameter can be a range. [10]

playlistmove {NAME} {FROM} {TO}

Moves the song at position FROM in the playlist NAME.m3u to the position TO.

rename {NAME} {NEW_NAME}

Renames the playlist NAME.m3u to NEW_NAME.m3u.

rm {NAME}

Removes the playlist NAME.m3u from the playlist directory.

save {NAME}

Saves the queue to NAME.m3u in the playlist directory.

The music database

albumart {URI} {OFFSET}

Locate album art for the given song and return a chunk of an album art image file at offset OFFSET.

This is currently implemented by searching the directory the file resides in for a file called cover.png, cover.jpg, cover.tiff or cover.bmp.

Returns the file size and actual number of bytes read at the requested offset, followed by the chunk requested as raw bytes (see Binary Responses), then a newline and the completion code.


albumart foo/bar.ogg 0
size: 1024768
binary: 8192
<8192 bytes>
count {FILTER} [group {GROUPTYPE}]

Count the number of songs and their total playtime in the database matching FILTER (see Filters). The following prints the number of songs whose title matches “Echoes”:

count title Echoes

The group keyword may be used to group the results by a tag. The first following example prints per-artist counts while the next prints the number of songs whose title matches “Echoes” grouped by artist:

count group artist
count title Echoes group artist

A group with an empty value contains counts of matching songs which don’t have this group tag. It exists only if at least one such song is found.

getfingerprint {URI}

Calculate the song’s audio fingerprint. Example (abbreviated fingerprint):

getfingerprint "foo/bar.ogg"
chromaprint: AQACcEmSREmWJJmkIT_6CCf64...

This command is only available if MPD was built with libchromaprint (-Dchromaprint=enabled).

find {FILTER} [sort {TYPE}] [window {START:END}]

Search the database for songs matching FILTER (see Filters).

sort sorts the result by the specified tag. The sort is descending if the tag is prefixed with a minus (‘-‘). Without sort, the order is undefined. Only the first tag value will be used, if multiple of the same type exist. To sort by “Artist”, “Album” or “AlbumArtist”, you should specify “ArtistSort”, “AlbumSort” or “AlbumArtistSort” instead. These will automatically fall back to the former if “*Sort” doesn’t exist. “AlbumArtist” falls back to just “Artist”. The type “Last-Modified” can sort by file modification time.

window can be used to query only a portion of the real response. The parameter is two zero-based record numbers; a start number and an end number.

findadd {FILTER} [sort {TYPE}] [window {START:END}] [position POS]

Search the database for songs matching FILTER (see Filters) and add them to the queue. Parameters have the same meaning as for find and searchadd.

list {TYPE} {FILTER} [group {GROUPTYPE}]

Lists unique tags values of the specified type. TYPE can be any tag supported by MPD.

Additional arguments may specify a filter. The group keyword may be used (repeatedly) to group the results by one or more tags.

The following example lists all album names, grouped by their respective (album) artist:

list album group albumartist

list file was implemented in an early MPD version, but does not appear to make a lot of sense. It still works (to avoid breaking compatibility), but is deprecated.

listall [URI]

Lists all songs and directories in URI.

Do not use this command. Do not manage a client-side copy of MPD’s database. That is fragile and adds huge overhead. It will break with large databases. Instead, query MPD whenever you need something.

listallinfo [URI]

Same as listall, except it also returns metadata info in the same format as lsinfo

Do not use this command. Do not manage a client-side copy of MPD’s database. That is fragile and adds huge overhead. It will break with large databases. Instead, query MPD whenever you need something.

listfiles {URI}

Lists the contents of the directory URI, including files are not recognized by MPD. URI can be a path relative to the music directory or an URI understood by one of the storage plugins. The response contains at least one line for each directory entry with the prefix “file: ” or “directory: “, and may be followed by file attributes such as “Last-Modified” and “size”.

For example, “smb://SERVER” returns a list of all shares on the given SMB/CIFS server; “nfs://servername/path” obtains a directory listing from the NFS server.

lsinfo [URI]

Lists the contents of the directory URI. The response contains records starting with file, directory or playlist, each followed by metadata (tags or other metadata).

When listing the root directory, this currently returns the list of stored playlists. This behavior is deprecated; use “listplaylists” instead.

This command may be used to list metadata of remote files (e.g. URI beginning with “http://” or “smb://”).

Clients that are connected via local socket may use this command to read the tags of an arbitrary local file (URI is an absolute path).

readcomments {URI}

Read “comments” (i.e. key-value pairs) from the file specified by “URI”. This “URI” can be a path relative to the music directory or an absolute path.

This command may be used to list metadata of remote files (e.g. URI beginning with “http://” or “smb://”).

The response consists of lines in the form “KEY: VALUE”. Comments with suspicious characters (e.g. newlines) are ignored silently.

The meaning of these depends on the codec, and not all decoder plugins support it. For example, on Ogg files, this lists the Vorbis comments.

readpicture {URI} {OFFSET}

Locate a picture for the given song and return a chunk of the image file at offset OFFSET. This is usually implemented by reading embedded pictures from binary tags (e.g. ID3v2’s APIC tag).

Returns the following values:

  • size: the total file size

  • type: the file’s MIME type (optional)

  • binary: see Binary Responses

If the song file was recognized, but there is no picture, the response is successful, but is otherwise empty.


readpicture foo/bar.ogg 0
size: 1024768
type: image/jpeg
binary: 8192
<8192 bytes>
searchadd {FILTER} [sort {TYPE}] [window {START:END}] [position POS]

Search the database for songs matching FILTER (see Filters) and add them to the queue.

Parameters have the same meaning as for search.

The position parameter specifies where the songs will be inserted. [8] It can be relative to the current song as in addid. [12]

searchaddpl {NAME} {FILTER} [sort {TYPE}] [window {START:END}] [position POS]

Search the database for songs matching FILTER (see Filters) and add them to the playlist named NAME.

If a playlist by that name doesn’t exist it is created.

Parameters have the same meaning as for search.

The position parameter specifies where the songs will be inserted. [11]

update [URI]

Updates the music database: find new files, remove deleted files, update modified files.

URI is a particular directory or song/file to update. If you do not specify it, everything is updated.

Prints updating_db: JOBID where JOBID is a positive number identifying the update job. You can read the current job id in the status response.

rescan [URI]

Same as update, but also rescans unmodified files.

Mounts and neighbors

A “storage” provides access to files in a directory tree. The most basic storage plugin is the “local” storage plugin which accesses the local file system, and there are plugins to access NFS and SMB servers.

Multiple storages can be “mounted” together, similar to the mount command on many operating systems, but without cooperation from the kernel. No superuser privileges are necessary, because this mapping exists only inside the MPD process.

mount {PATH} {URI}

Mount the specified remote storage URI at the given path. Example:

mount foo nfs://
unmount {PATH}

Unmounts the specified path. Example:

unmount foo

Queries a list of all mounts. By default, this contains just the configured music_directory. Example:

storage: /home/foo/music
mount: foo
storage: nfs://

Queries a list of “neighbors” (e.g. accessible file servers on the local net). Items on that list may be used with the mount command. Example:

neighbor: smb://FOO
name: FOO (Samba 4.1.11-Debian)


“Stickers” [2] are pieces of information attached to existing MPD objects (e.g. song files, directories, albums; but currently, they are only implemented for song). Clients can create arbitrary name/value pairs. MPD itself does not assume any special meaning in them.

The goal is to allow clients to share additional (possibly dynamic) information about songs, which is neither stored on the client (not available to other clients), nor stored in the song files (MPD has no write access).

Client developers should create a standard for common sticker names, to ensure interoperability.

Objects which may have stickers are addressed by their object type (“song” for song objects) and their URI (the path within the database for songs).

sticker get {TYPE} {URI} {NAME}

Reads a sticker value for the specified object.

sticker set {TYPE} {URI} {NAME} {VALUE}

Adds a sticker value to the specified object. If a sticker item with that name already exists, it is replaced.

sticker delete {TYPE} {URI} [NAME]

Deletes a sticker value from the specified object. If you do not specify a sticker name, all sticker values are deleted.

sticker list {TYPE} {URI}

Lists the stickers for the specified object.

sticker find {TYPE} {URI} {NAME}

Searches the sticker database for stickers with the specified name, below the specified directory (URI). For each matching song, it prints the URI and that one sticker’s value.

sticker find {TYPE} {URI} {NAME} = {VALUE}

Searches for stickers with the given value.

Other supported operators are: “<”, “>

Connection settings


Closes the connection to MPD. MPD will try to send the remaining output buffer before it actually closes the connection, but that cannot be guaranteed. This command will not generate a response.

Clients should not use this command; instead, they should just close the socket.


Kills MPD.

Do not use this command. Send SIGTERM to MPD instead, or better: let your service manager handle MPD shutdown (e.g. systemctl stop mpd).

password {PASSWORD}

This is used for authentication with the server. PASSWORD is simply the plaintext password.


Does nothing but return “OK”.

binarylimit SIZE [7]

Set the maximum binary response size for the current connection to the specified number of bytes.

A bigger value means less overhead for transmitting large entities, but it also means that the connection is blocked for a longer time.


Shows a list of available tag types. It is an intersection of the metadata_to_use setting and this client’s tag mask.

About the tag mask: each client can decide to disable any number of tag types, which will be omitted from responses to this client. That is a good idea, because it makes responses smaller. The following tagtypes sub commands configure this list.

tagtypes disable {NAME...}

Remove one or more tags from the list of tag types the client is interested in. These will be omitted from responses to this client.

tagtypes enable {NAME...}

Re-enable one or more tags from the list of tag types for this client. These will no longer be hidden from responses to this client.

tagtypes clear

Clear the list of tag types this client is interested in. This means that MPD will not send any tags to this client.

tagtypes all

Announce that this client is interested in all tag types. This is the default setting for new clients.

Partition commands

These commands allow a client to inspect and manage “partitions”. A partition is one frontend of a multi-player MPD process: it has separate queue, player and outputs. A client is assigned to one partition at a time.

partition {NAME}

Switch the client to a different partition.


Print a list of partitions. Each partition starts with a partition keyword and the partition’s name, followed by information about the partition.

newpartition {NAME}

Create a new partition.

delpartition {NAME}

Delete a partition. The partition must be empty (no connected clients and no outputs).

moveoutput {OUTPUTNAME}

Move an output to the current partition.

Audio output devices

disableoutput {ID}

Turns an output off.

enableoutput {ID}

Turns an output on.

toggleoutput {ID}

Turns an output on or off, depending on the current state.


Shows information about all outputs.

outputid: 0
outputname: My ALSA Device
plugin: alsa
outputenabled: 0
attribute: dop=0

Return information:

  • outputid: ID of the output. May change between executions

  • outputname: Name of the output. It can be any.

  • outputenabled: Status of the output. 0 if disabled, 1 if enabled.

outputset {ID} {NAME} {VALUE}

Set a runtime attribute. These are specific to the output plugin, and supported values are usually printed in the outputs response.



Dumps configuration values that may be interesting for the client. This command is only permitted to “local” clients (connected via local socket).

The following response attributes are available:

  • music_directory: The absolute path of the music directory.


Shows which commands the current user has access to.


Shows which commands the current user does not have access to.


Gets a list of available URL handlers.


Print a list of decoder plugins, followed by their supported suffixes and MIME types. Example response:

plugin: mad
suffix: mp3
suffix: mp2
mime_type: audio/mpeg
plugin: mpcdec
suffix: mpc

Client to client

Clients can communicate with each others over “channels”. A channel is created by a client subscribing to it. More than one client can be subscribed to a channel at a time; all of them will receive the messages which get sent to it.

Each time a client subscribes or unsubscribes, the global idle event subscription is generated. In conjunction with the channels command, this may be used to auto-detect clients providing additional services.

New messages are indicated by the message idle event.

If your MPD instance has multiple partitions, note that client-to-client messages are local to the current partition.

subscribe {NAME}

Subscribe to a channel. The channel is created if it does not exist already. The name may consist of alphanumeric ASCII characters plus underscore, dash, dot and colon.

unsubscribe {NAME}

Unsubscribe from a channel.


Obtain a list of all channels. The response is a list of “channel:” lines.


Reads messages for this client. The response is a list of “channel:” and “message:” lines.

sendmessage {CHANNEL} {TEXT}

Send a message to the specified channel.