diff --git a/troi/listenbrainz/recs.py b/troi/listenbrainz/recs.py
index 9ff8a242..945d4f51 100644
--- a/troi/listenbrainz/recs.py
+++ b/troi/listenbrainz/recs.py
@@ -22,8 +22,6 @@ class UserRecordingRecommendationsElement(Element):
     def __init__(self, user_name, artist_type, count=25, offset=0):
         super().__init__()
         self.client = liblistenbrainz.ListenBrainz()
-        if auth_token:
-            self.client.set_auth_token(auth_token)
         self.user_name = user_name
         self.count = count
         self.offset = offset
diff --git a/troi/recording_search_service.py b/troi/recording_search_service.py
new file mode 100644
index 00000000..52db1154
--- /dev/null
+++ b/troi/recording_search_service.py
@@ -0,0 +1,88 @@
+from abc import abstractmethod
+
+import requests
+
+from troi import Recording, Artist
+from troi.service import Service
+from troi.splitter import plist
+
+# NOTES FOR LB API improvements:
+# Tags:
+#    - flatten tag data for simplicity
+#    - use series of tighter random spots to speed up or searches for popular tags
+# Artist:
+#    - Support percent based popular track lookups and move logic to server
+
+
+class RecordingSearchByTagService(Service):
+
+    SLUG = "recording-search-by-tag"
+
+    def __init__(self):
+        super().__init__(self.SLUG)
+
+    def search(self, tags, operator, begin_percent, end_percent, num_recordings):
+        """
+            Fetch the tag data from the LB API and return it as a dict.
+        """
+
+        data = {
+            "condition": operator,
+            "count": num_recordings,
+            "begin_percent": begin_percent,
+            "end_percent": end_percent,
+            "tag": tags,
+            "min_tag_count": 1
+        }
+        r = requests.get("https://api.listenbrainz.org/1/lb-radio/tags", params=data)
+        if r.status_code != 200:
+            raise RuntimeError(f"Cannot fetch recordings for tags. {r.text}")
+
+        recordings = []
+        for rec in self.flatten_tag_data(dict(r.json())):
+            recordings.append(Recording(mbid=rec["recording_mbid"]))
+
+        return plist(recordings)
+
+    def flatten_tag_data(self, tag_data):
+
+        flat_data = list(tag_data["recording"])
+        flat_data.extend(list(tag_data["release-group"]))
+        flat_data.extend(list(tag_data["artist"]))
+
+        return sorted(flat_data, key=lambda f: f["percent"], reverse=True)
+
+
+class RecordingSearchByArtistService(Service):
+
+    SLUG = "recording-search-by-artist"
+
+    def __init__(self):
+        super().__init__(self.SLUG)
+
+    def search(self, artists, begin_percent, end_percent, num_recordings):
+        """
+            Fetch the artist data from the LB API and return it as a dict.
+
+            NOTE: This search is poor -- it should span all recordings by an artist not, just the top ones!
+        """
+
+        artists_recordings = {}
+        for artist_mbid in artists:
+            params={"artist_mbid": artist_mbid}
+            r = requests.get("https://api.listenbrainz.org/1/popularity/top-recordings-for-artist", params={"artist_mbid": artist_mbid})
+            if r.status_code != 200:
+                raise RuntimeError(f"Cannot fetch top artist recordings: {r.status_code} ({r.text})")
+
+            recordings = plist()
+            for recording in r.json():
+                artist = Artist(mbids=recording["artist_mbids"], name=recording["artist_name"])
+                recordings.append(
+                    Recording(mbid=recording["recording_mbid"],
+                              name=recording["recording_name"],
+                              duration=recording["length"],
+                              artist=artist))
+
+            artists_recordings[artist_mbid] = recordings.random_item(begin_percent, end_percent, num_recordings)
+
+        return artists_recordings
diff --git a/troi/service.py b/troi/service.py
new file mode 100755
index 00000000..097f1058
--- /dev/null
+++ b/troi/service.py
@@ -0,0 +1,18 @@
+class Service:
+    """
+        Services allow troi users to provide local versions of key search functions
+        so that both global (e.g. MusicBrainz) and local search contexts can be used.
+        
+        Downstream users will be able to provide localized version of search functions
+        to provide lists of Recordings that meet certain requirements for playlist inclusion.
+
+        Only one service can be registered for any given service slug at a time. The most
+        recently registered service will be use for the next playlist generation.
+    """
+
+    def __init__(self, slug):
+        self._slug = slug
+
+    @property
+    def slug(self):
+        return self._slug