diff --git a/src/mpd_now_playable/mpd/listener.py b/src/mpd_now_playable/mpd/listener.py
index f787223..50f256a 100644
--- a/src/mpd_now_playable/mpd/listener.py
+++ b/src/mpd_now_playable/mpd/listener.py
@@ -1,7 +1,6 @@
 import asyncio
 from collections.abc import Iterable
 from pathlib import Path
-from uuid import UUID
 
 from mpd.asyncio import MPDClient
 from mpd.base import CommandError
@@ -10,7 +9,7 @@ from yarl import URL
 
 from ..config.model import MpdConfig
 from ..player import Player
-from ..song import Artwork, PlaybackState, Song, to_artwork
+from ..song import Artwork, PlaybackState, Song, to_artwork, to_brainz
 from ..song_receiver import Receiver
 from ..tools.types import convert_if_exists, un_maybe_plural
 from .artwork_cache import MpdArtworkCache
@@ -25,10 +24,6 @@ def mpd_current_to_song(
 		queue_index=int(current["pos"]),
 		queue_length=int(status["playlistlength"]),
 		file=Path(current["file"]),
-		musicbrainz_trackid=convert_if_exists(current.get("musicbrainz_trackid"), UUID),
-		musicbrainz_releasetrackid=convert_if_exists(
-			current.get("musicbrainz_releasetrackid"), UUID
-		),
 		title=current.get("title"),
 		artist=un_maybe_plural(current.get("artist")),
 		album=un_maybe_plural(current.get("album")),
@@ -39,6 +34,7 @@ def mpd_current_to_song(
 		disc=convert_if_exists(current.get("disc"), int),
 		duration=float(status["duration"]),
 		elapsed=float(status["elapsed"]),
+		musicbrainz=to_brainz(current),
 		art=art,
 	)
 
diff --git a/src/mpd_now_playable/receivers/cocoa/persistent_id.py b/src/mpd_now_playable/receivers/cocoa/persistent_id.py
index 0ce50a1..12c992d 100644
--- a/src/mpd_now_playable/receivers/cocoa/persistent_id.py
+++ b/src/mpd_now_playable/receivers/cocoa/persistent_id.py
@@ -7,25 +7,27 @@ from ...song import Song
 
 # The maximum size for a BLAKE2b "person" value is sixteen bytes, so we need to be concise.
 HASH_PERSON_PREFIX: Final = b"mnp.mac."
-TRACKID_HASH_PERSON: Final = HASH_PERSON_PREFIX + b"mb_tid"
-RELEASETRACKID_HASH_PERSON: Final = HASH_PERSON_PREFIX + b"mb_rtid"
+RECORDING_ID_HASH_PERSON: Final = HASH_PERSON_PREFIX + b"mb_rid"
+TRACK_ID_HASH_PERSON: Final = HASH_PERSON_PREFIX + b"mb_tid"
 FILE_HASH_PERSON: Final = HASH_PERSON_PREFIX + b"f"
 
 PERSISTENT_ID_BITS: Final = 64
 PERSISTENT_ID_BYTES: Final = PERSISTENT_ID_BITS // 8
 
 
-def digest_trackid(trackid: UUID) -> bytes:
+def digest_recording_id(recording_id: UUID) -> bytes:
 	return blake2b(
-		trackid.bytes, digest_size=PERSISTENT_ID_BYTES, person=TRACKID_HASH_PERSON
+		recording_id.bytes,
+		digest_size=PERSISTENT_ID_BYTES,
+		person=RECORDING_ID_HASH_PERSON,
 	).digest()
 
 
-def digest_releasetrackid(trackid: UUID) -> bytes:
+def digest_track_id(track_id: UUID) -> bytes:
 	return blake2b(
-		trackid.bytes,
+		track_id.bytes,
 		digest_size=PERSISTENT_ID_BYTES,
-		person=RELEASETRACKID_HASH_PERSON,
+		person=TRACK_ID_HASH_PERSON,
 	).digest()
 
 
@@ -41,10 +43,10 @@ def digest_file_uri(file: Path) -> bytes:
 # that from the file URI. BLAKE2 can be customised to different digest sizes,
 # making it perfect for this problem.
 def song_to_persistent_id(song: Song) -> int:
-	if song.musicbrainz_trackid:
-		hashed_id = digest_trackid(song.musicbrainz_trackid)
-	elif song.musicbrainz_releasetrackid:
-		hashed_id = digest_releasetrackid(song.musicbrainz_releasetrackid)
+	if song.musicbrainz.recording:
+		hashed_id = digest_recording_id(song.musicbrainz.recording)
+	elif song.musicbrainz.track:
+		hashed_id = digest_track_id(song.musicbrainz.track)
 	else:
 		hashed_id = digest_file_uri(song.file)
 	return int.from_bytes(hashed_id)
diff --git a/src/mpd_now_playable/song.py b/src/mpd_now_playable/song.py
deleted file mode 100644
index 487c9cb..0000000
--- a/src/mpd_now_playable/song.py
+++ /dev/null
@@ -1,55 +0,0 @@
-from dataclasses import dataclass, field
-from enum import StrEnum
-from pathlib import Path
-from typing import Literal
-from uuid import UUID
-
-from pydantic.type_adapter import TypeAdapter
-
-
-class PlaybackState(StrEnum):
-	play = "play"
-	pause = "pause"
-	stop = "stop"
-
-
-@dataclass(slots=True)
-class HasArtwork:
-	data: bytes = field(repr=False)
-
-
-@dataclass(slots=True)
-class NoArtwork:
-	def __bool__(self) -> Literal[False]:
-		return False
-
-
-Artwork = HasArtwork | NoArtwork
-ArtworkSchema: TypeAdapter[Artwork] = TypeAdapter(HasArtwork | NoArtwork)
-
-
-def to_artwork(art: bytes | None) -> Artwork:
-	if art is None:
-		return NoArtwork()
-	return HasArtwork(art)
-
-
-@dataclass(slots=True)
-class Song:
-	state: PlaybackState
-	queue_index: int
-	queue_length: int
-	file: Path
-	musicbrainz_trackid: UUID | None
-	musicbrainz_releasetrackid: UUID | None
-	title: str | None
-	artist: list[str]
-	composer: list[str]
-	album: list[str]
-	album_artist: list[str]
-	track: int | None
-	disc: int | None
-	genre: list[str]
-	duration: float
-	elapsed: float
-	art: Artwork
diff --git a/src/mpd_now_playable/song/__init__.py b/src/mpd_now_playable/song/__init__.py
new file mode 100644
index 0000000..80e5f4f
--- /dev/null
+++ b/src/mpd_now_playable/song/__init__.py
@@ -0,0 +1,12 @@
+from .artwork import Artwork, ArtworkSchema, to_artwork
+from .musicbrainz import to_brainz
+from .song import PlaybackState, Song
+
+__all__ = (
+	"Artwork",
+	"ArtworkSchema",
+	"to_artwork",
+	"to_brainz",
+	"PlaybackState",
+	"Song",
+)
diff --git a/src/mpd_now_playable/song/artwork.py b/src/mpd_now_playable/song/artwork.py
new file mode 100644
index 0000000..4c4d941
--- /dev/null
+++ b/src/mpd_now_playable/song/artwork.py
@@ -0,0 +1,25 @@
+from dataclasses import dataclass, field
+from typing import Literal
+
+from pydantic.type_adapter import TypeAdapter
+
+
+@dataclass(slots=True)
+class HasArtwork:
+	data: bytes = field(repr=False)
+
+
+@dataclass(slots=True)
+class NoArtwork:
+	def __bool__(self) -> Literal[False]:
+		return False
+
+
+Artwork = HasArtwork | NoArtwork
+ArtworkSchema: TypeAdapter[Artwork] = TypeAdapter(HasArtwork | NoArtwork)
+
+
+def to_artwork(art: bytes | None) -> Artwork:
+	if art is None:
+		return NoArtwork()
+	return HasArtwork(art)
diff --git a/src/mpd_now_playable/song/musicbrainz.py b/src/mpd_now_playable/song/musicbrainz.py
new file mode 100644
index 0000000..7478d2c
--- /dev/null
+++ b/src/mpd_now_playable/song/musicbrainz.py
@@ -0,0 +1,96 @@
+from dataclasses import dataclass
+from functools import partial
+from typing import TypedDict
+from uuid import UUID
+
+from ..tools.types import option_fmap
+
+option_uuid = partial(option_fmap, UUID)
+
+
+class MusicBrainzTags(TypedDict, total=False):
+	"""
+	The MusicBrainz tags mpd-now-playable expects and will load (all optional).
+	They're named slightly differently than the actual MusicBrainz IDs they
+	store - use to_brainz to map them across to their canonical form.
+	https://picard-docs.musicbrainz.org/downloads/MusicBrainz_Picard_Tag_Map.html
+	"""
+
+	#: MusicBrainz Recording ID
+	musicbrainz_trackid: str
+	#: MusicBrainz Track ID
+	musicbrainz_releasetrackid: str
+	#: MusicBrainz Artist ID
+	musicbrainz_artistid: str
+	#: MusicBrainz Release ID
+	musicbrainz_albumid: str
+	#: MusicBrainz Release Artist ID
+	musicbrainz_albumartistid: str
+	#: MusicBrainz Release Group ID
+	musicbrainz_releasegroupid: str
+	#: MusicBrainz Work ID
+	musicbrainz_workid: str
+
+
+@dataclass(slots=True)
+class MusicBrainzIds:
+	#: A MusicBrainz recording represents audio from a specific performance.
+	#: For example, if the same song was released as a studio recording and as
+	#: a live performance, those two versions of the song are different
+	#: recordings. The song itself is considered a "work", of which two
+	#: recordings were made. However, recordings are not always associated with
+	#: a work in the MusicBrainz database, and Picard won't load work IDs by
+	#: default (you have to enable "use track relationships" in the options),
+	#: so recording IDs are a much more reliable way to identify a particular
+	#: song.
+	#: https://musicbrainz.org/doc/Recording
+	recording: UUID | None
+
+	#: A MusicBrainz work represents the idea of a particular song or creation
+	#: (it doesn't have to be audio). Each work may have multiple recordings
+	#: (studio versus live, different performers, etc.), with the work ID
+	#: grouping them together.
+	#: https://musicbrainz.org/doc/Work
+	work: UUID | None
+
+	#: A MusicBrainz track represents a specific instance of a recording
+	#: appearing as part of some release. For example, if the same song appears
+	#: on both two-CD and four-CD versions of a soundtrack, then it will be
+	#: considered the same "recording" in both cases, but different "tracks".
+	#: https://musicbrainz.org/doc/Track
+	track: UUID | None
+
+	#: https://musicbrainz.org/doc/Artist
+	artist: UUID | None
+
+	#: A MusicBrainz release roughly corresponds to an "album", and indeed is
+	#: stored in a tag called MUSICBRAINZ_ALBUMID. The more general name is
+	#: meant to encompass all the different ways music can be released.
+	#: https://musicbrainz.org/doc/Release
+	release: UUID | None
+
+	#: Again, the release artist corresponds to an "album artist". These MBIDs
+	#: refer to the same artists in the MusicBrainz database that individual
+	#: recordings' artist MBIDs do.
+	release_artist: UUID | None
+
+	#: A MusicBrainz release group roughly corresponds to "all the editions of
+	#: a particular album". For example, if the same album were released on CD,
+	#: vinyl records, and as a digital download, then all of those would be
+	#: different releases but share a release group. Note that MPD's support
+	#: for this tag is relatively new (July 2023) and doesn't seem especially
+	#: reliable, so it might be missing here even if your music has been tagged
+	#: with it. Not sure why. https://musicbrainz.org/doc/Release_Group
+	release_group: UUID | None
+
+
+def to_brainz(tags: MusicBrainzTags) -> MusicBrainzIds:
+	return MusicBrainzIds(
+		recording=option_uuid(tags.get("musicbrainz_trackid")),
+		work=option_uuid(tags.get("musicbrainz_workid")),
+		track=option_uuid(tags.get("musicbrainz_releasetrackid")),
+		artist=option_uuid(tags.get("musicbrainz_artistid")),
+		release=option_uuid(tags.get("musicbrainz_albumid")),
+		release_artist=option_uuid(tags.get("musicbrainz_albumartistid")),
+		release_group=option_uuid(tags.get("musicbrainz_releasegroupid")),
+	)
diff --git a/src/mpd_now_playable/song/song.py b/src/mpd_now_playable/song/song.py
new file mode 100644
index 0000000..e30c9ca
--- /dev/null
+++ b/src/mpd_now_playable/song/song.py
@@ -0,0 +1,32 @@
+from dataclasses import dataclass
+from enum import StrEnum
+from pathlib import Path
+
+from .artwork import Artwork
+from .musicbrainz import MusicBrainzIds
+
+
+class PlaybackState(StrEnum):
+	play = "play"
+	pause = "pause"
+	stop = "stop"
+
+
+@dataclass(slots=True)
+class Song:
+	state: PlaybackState
+	queue_index: int
+	queue_length: int
+	file: Path
+	title: str | None
+	artist: list[str]
+	composer: list[str]
+	album: list[str]
+	album_artist: list[str]
+	track: int | None
+	disc: int | None
+	genre: list[str]
+	duration: float
+	elapsed: float
+	art: Artwork
+	musicbrainz: MusicBrainzIds
diff --git a/src/mpd_now_playable/tools/types.py b/src/mpd_now_playable/tools/types.py
index 023641d..da62c90 100644
--- a/src/mpd_now_playable/tools/types.py
+++ b/src/mpd_now_playable/tools/types.py
@@ -4,6 +4,7 @@ from typing import Any, TypeAlias, TypeVar
 __all__ = (
 	"AnyExceptList",
 	"MaybePlural",
+	"option_fmap",
 	"convert_if_exists",
 	"un_maybe_plural",
 )
@@ -29,6 +30,7 @@ AnyExceptList = (
 
 
 U = TypeVar("U")
+V = TypeVar("V")
 
 
 def not_none(value: U | None) -> U:
@@ -37,6 +39,12 @@ def not_none(value: U | None) -> U:
 	return value
 
 
+def option_fmap(f: Callable[[U], V], value: U | None) -> V | None:
+	if value is None:
+		return None
+	return f(value)
+
+
 def convert_if_exists(value: str | None, converter: Callable[[str], U]) -> U | None:
 	if value is None:
 		return None