diff --git a/zephyr/lib/event_queue.py b/zephyr/lib/event_queue.py
new file mode 100644
index 0000000000..69c3a11d08
--- /dev/null
+++ b/zephyr/lib/event_queue.py
@@ -0,0 +1,120 @@
+from django.conf import settings
+from collections import deque
+from tornado.ioloop import PeriodicCallback
+import time
+import socket
+import logging
+import simplejson
+
+IDLE_EVENT_QUEUE_TIMEOUT_SECS = 60 * 10
+
+class ClientDescriptor(object):
+    def __init__(self, user_profile_id, id, apply_markdown=True):
+        self.user_profile_id = user_profile_id
+        self.current_handler = None
+        self.event_queue = EventQueue(id)
+        self.last_connection_time = time.time()
+        self.apply_markdown = apply_markdown
+
+    def add_event(self, event):
+        if self.current_handler is not None:
+            self.current_handler._request._time_restarted = time.time()
+
+        self.event_queue.push(event)
+        self.check_connection()
+        if self.current_handler is not None:
+            try:
+                self.current_handler.humbug_finish(dict(result='success', msg='',
+                                                        events=[event],
+                                                        queue_id=self.event_queue.id),
+                                                   self.current_handler._request,
+                                                   apply_markdown=self.apply_markdown)
+                return
+            except socket.error:
+                pass
+
+    def idle(self, now):
+        self.check_connection()
+        return (self.current_handler is None
+                and now - self.last_connection_time >= IDLE_EVENT_QUEUE_TIMEOUT_SECS)
+
+    def connect_handler(self, handler):
+        self.current_handler = handler
+        self.last_connection_time = time.time()
+
+    def disconnect_handler(self):
+        self.current_handler = None
+
+    def check_connection(self):
+        if (self.current_handler is not None
+            and self.current_handler.request.connection.stream.closed()):
+            self.current_handler = None
+
+class EventQueue(object):
+    def __init__(self, id):
+        self.queue = deque()
+        self.connected_handler = None
+        self.next_event_id = 0
+        self.id = id
+
+    def push(self, event):
+        event['id'] = self.next_event_id
+        self.next_event_id += 1
+        self.queue.append(event)
+
+    def pop(self):
+        return self.queue.popleft()
+
+    def empty(self):
+        return len(self.queue) == 0
+
+    def prune(self, through_id):
+        while not self.empty() and self.queue[0]['id'] <= through_id:
+            self.pop()
+
+    def contents(self):
+        return list(self.queue)
+
+# maps queue ids to client descriptors
+clients = {}
+# maps user id to list of client descriptors
+user_clients = {}
+
+next_queue_id = 0
+
+def allocate_client_descriptor(user_profile_id, apply_markdown):
+    global next_queue_id
+    id = str(settings.SERVER_GENERATION) + ':' + str(next_queue_id)
+    next_queue_id += 1
+    client = ClientDescriptor(user_profile_id, id, apply_markdown)
+    clients[id] = client
+    user_clients.setdefault(user_profile_id, []).append(client)
+    return client
+
+EVENT_QUEUE_GC_FREQ_MSECS = 1000 * 60 * 5
+
+def gc_event_queues():
+    start = time.time()
+    to_remove = set()
+    affected_users = set()
+    for (id, client) in clients.iteritems():
+        if client.idle(start):
+            to_remove.add(id)
+            affected_users.add(client.user_profile_id)
+
+    for id in to_remove:
+        del clients[id]
+
+    for user_id in affected_users:
+        new_client_list = filter(lambda c: c.event_queue.id not in to_remove,
+                                user_clients[user_id])
+        user_clients[user_id] = new_client_list
+
+    logging.info(('Tornado removed %d idle event queues owned by %d users in %.3fs.'
+                  + '  Now %d active queues')
+                 % (len(to_remove), len(affected_users), time.time() - start,
+                    len(clients)))
+
+def setup_event_queue_gc(io_loop):
+    pc = PeriodicCallback(gc_event_queues, EVENT_QUEUE_GC_FREQ_MSECS, io_loop)
+    pc.start()
diff --git a/zephyr/management/commands/runtornado.py b/zephyr/management/commands/runtornado.py
index cf86e0b392..bf2cc33a61 100644
--- a/zephyr/management/commands/runtornado.py
+++ b/zephyr/management/commands/runtornado.py
@@ -19,6 +19,7 @@ from tornado import ioloop
 from zephyr.lib.debug import interactive_debug_listen
 from zephyr.lib.response import json_response
 from zephyr import tornado_callbacks
+from zephyr.lib.event_queue import setup_event_queue_gc
 
 if settings.USING_RABBITMQ:
     from zephyr.lib.queue import queue_client
@@ -102,6 +103,7 @@ class Command(BaseCommand):
                 if django.conf.settings.DEBUG:
                     ioloop.IOLoop.instance().set_blocking_log_threshold(5)
 
+                setup_event_queue_gc(ioloop.IOLoop.instance())
                 ioloop.IOLoop.instance().start()
             except KeyboardInterrupt:
                 sys.exit(0)
diff --git a/zephyr/tornado_callbacks.py b/zephyr/tornado_callbacks.py
index f82ec377fb..207eb14348 100644
--- a/zephyr/tornado_callbacks.py
+++ b/zephyr/tornado_callbacks.py
@@ -5,6 +5,7 @@ from zephyr.models import Message, UserProfile, UserMessage, \
 from zephyr.decorator import JsonableError
 from zephyr.lib.cache_helpers import cache_get_message
 from zephyr.lib.queue import queue_json_publish
+from zephyr.lib.event_queue import user_clients
 
 import os
 import sys
@@ -239,12 +240,21 @@ def update_pointer(user_profile_id, new_pointer):
                          new_pointer=new_pointer,
                          update_types=["pointer_update"])
 
+    for client in user_clients.get(user_profile_id, []):
+        event = dict(type='pointer', pointer=new_pointer)
+        client.add_event(event)
+
 def process_new_message(data):
     message = cache_get_message(data['message'])
 
     for user_profile_id in data['users']:
         user_receive_message(user_profile_id, message)
 
+        for client in user_clients.get(user_profile_id, []):
+            event = dict(type='message', message=message.to_dict(client.apply_markdown))
+            client.add_event(event)
+
+
     if 'stream_name' in data:
         stream_receive_message(data['realm_id'], data['stream_name'], message)
 
diff --git a/zephyr/tornadoviews.py b/zephyr/tornadoviews.py
index 0f65fe78d9..e503a448d9 100644
--- a/zephyr/tornadoviews.py
+++ b/zephyr/tornadoviews.py
@@ -6,7 +6,7 @@ from zephyr.decorator import asynchronous, authenticated_api_view, \
     has_request_variables, POST, to_non_negative_int, json_to_bool, \
     JsonableError, authenticated_rest_api_view, REQ
 
-from zephyr.lib.response import json_success, json_error
+from zephyr.lib.response import json_response, json_success, json_error
 
 from zephyr.tornado_callbacks import \
     get_user_pointer, fetch_stream_messages, fetch_user_messages, \
@@ -14,6 +14,7 @@ from zephyr.tornado_callbacks import \
     add_pointer_update_callback, process_notification
 
 from zephyr.lib.cache_helpers import cache_get_message
+from zephyr.lib.event_queue import allocate_client_descriptor, clients
 
 import datetime
 import simplejson
@@ -192,3 +193,44 @@ def get_updates_backend(request, user_profile, handler, client_id,
 
     # runtornado recognizes this special return value.
     return RespondAsynchronously
+
+@asynchronous
+@authenticated_json_post_view
+def json_get_events(request, user_profile, handler):
+    return get_events_backend(request, user_profile, handler)
+
+@asynchronous
+@authenticated_rest_api_view
+@has_request_variables
+def rest_get_events(request, user_profile, handler,
+                    apply_markdown=REQ(default=False, converter=json_to_bool)):
+    return get_events_backend(request, user_profile, handler,
+                              apply_markdown=apply_markdown)
+
+@has_request_variables
+def get_events_backend(request, user_profile, handler,
+                       last_event_id = REQ(converter=to_non_negative_int, default=None),
+                       queue_id = REQ(default=None), apply_markdown=True):
+    if queue_id is None:
+        client = allocate_client_descriptor(user_profile.id, apply_markdown)
+        queue_id = client.event_queue.id
+    else:
+        if last_event_id is None:
+            return json_error("Missing 'last_event_id' argument")
+        client = clients.get(queue_id)
+        if client is None:
+            return json_error("Bad event queue id: %s" % (queue_id,))
+        if user_profile.id != client.user_profile_id:
+            return json_error("You are not authorized to get events from this queue")
+        client.event_queue.prune(last_event_id)
+        client.disconnect_handler()
+
+    if not client.event_queue.empty():
+        return json_success({'events': client.event_queue.contents(),
+                             'queue_id': queue_id})
+
+    handler._request = request
+    client.connect_handler(handler)
+
+    # runtornado recognizes this special return value.
+    return RespondAsynchronously