diff --git a/README.rst b/README.rst
index bb28936..253b23a 100644
--- a/README.rst
+++ b/README.rst
@@ -68,7 +68,6 @@ Roadmap
 =======
 
 - test a few more performance improvement ideas
-- handle the mempool
 - implement light caching of client responses
 - yield during expensive requests and/or penalize the connection
 - improve DB abstraction so LMDB is not penalized
diff --git a/RELEASE-NOTES b/RELEASE-NOTES
index 6ffc948..eba9cdf 100644
--- a/RELEASE-NOTES
+++ b/RELEASE-NOTES
@@ -1,3 +1,14 @@
+Version 0.06
+------------
+
+- mempool support.  ElectrumX maintains a representation of the daemon's
+  mempool and serves unconfirmed transactions and balances to clients.
+
+Version 0.05
+------------
+
+- fixed a bug in 0.04 that stopped ElectrumX serving once synced
+
 Version 0.04
 ------------
 
diff --git a/samples/scripts/NOTES b/samples/scripts/NOTES
index 7da8c62..e56c4ec 100644
--- a/samples/scripts/NOTES
+++ b/samples/scripts/NOTES
@@ -22,8 +22,10 @@ REORG_LIMIT   - maximum number of blocks to be able to handle in a chain
                 reorganisation.  ElectrumX retains some fairly compact
                 undo information for this many blocks in levelDB.
                 Default is 200.
-TCP_PORT      - if set will serve Electrum clients on that port
-SSL_PORT      - if set will serve Electrum clients over SSL on that port.
+HOST          - the host that the TCP and SSL servers will use.  Defaults to
+                localhost.
+TCP_PORT      - if set will serve Electrum TCP clients on that HOST:TCP_PORT
+SSL_PORT      - if set will serve Electrum SSL clients on that HOST:SSL_PORT
                 If set, SSL_CERTFILE and SSL_KEYFILE must be filesystem paths.
 RPC_PORT      - Listen on this port for local RPC connections, defaults to
                 8000.
diff --git a/server/block_processor.py b/server/block_processor.py
index 742b536..34c9bce 100644
--- a/server/block_processor.py
+++ b/server/block_processor.py
@@ -21,6 +21,7 @@ from server.cache import FSCache, UTXOCache, NO_CACHE_ENTRY
 from server.daemon import DaemonError
 from lib.hash import hash_to_str
 from lib.script import ScriptPubKey
+from lib.tx import Deserializer
 from lib.util import chunks, LoggedClass
 from server.storage import open_db
 
@@ -50,7 +51,7 @@ class Prefetcher(LoggedClass):
         self.queue = asyncio.Queue()
         self.queue_size = 0
         self.fetched_height = height
-        self.mempool = []
+        self.mempool_hashes = []
         # Target cache size.  Has little effect on sync time.
         self.target_cache_size = 10 * 1024 * 1024
         # First fetch to be 10 blocks
@@ -74,7 +75,7 @@ class Prefetcher(LoggedClass):
         blocks, height, size = await self.queue.get()
         self.queue_size -= size
         if height == self.daemon.cached_height():
-            return blocks, self.mempool
+            return blocks, self.mempool_hashes
         else:
             return blocks, None
 
@@ -99,7 +100,7 @@ class Prefetcher(LoggedClass):
             self.fetched_height += len(blocks)
             caught_up = self.fetched_height == self.daemon.cached_height()
             if caught_up:
-                self.mempool = await self.daemon.mempool_hashes()
+                self.mempool_hashes = await self.daemon.mempool_hashes()
 
             # Wake up block processor if we have something
             if blocks or caught_up:
@@ -137,6 +138,143 @@ class Prefetcher(LoggedClass):
 
         return blocks, size
 
+class MissingUTXOError(Exception):
+    pass
+
+class MemPool(LoggedClass):
+    '''Representation of the daemon's mempool.
+
+    Updated regularly in caught-up state.  Goal is to enable efficient
+    response to the value() and transactions() calls.
+
+    To that end we maintain the following maps:
+
+       tx_hash -> [txin_pairs, txout_pairs, unconfirmed]
+       hash168 -> set of all tx hashes in which the hash168 appears
+
+    A pair is a (hash168, value) tuple.  Unconfirmed is true if any of the
+    tx's txins are unconfirmed.  tx hashes are hex strings.
+    '''
+
+    def __init__(self, bp):
+        super().__init__()
+        self.txs = {}
+        self.hash168s = defaultdict(set)  # None can be a key
+        self.bp = bp
+
+    async def update(self, hex_hashes):
+        '''Update state given the current mempool to the passed set of hashes.
+
+        Remove transactions that are no longer in our mempool.
+        Request new transactions we don't have then add to our mempool.
+        '''
+        hex_hashes = set(hex_hashes)
+        touched = set()
+
+        if not self.txs:
+            self.logger.info('initial fetch of {:,d} daemon mempool txs'
+                             .format(len(hex_hashes)))
+
+        # Remove gone items
+        gone = set(self.txs).difference(hex_hashes)
+        for hex_hash in gone:
+            txin_pairs, txout_pairs, unconfirmed = self.txs.pop(hex_hash)
+            hash168s = set(hash168 for hash168, value in txin_pairs)
+            hash168s.update(hash168 for hash168, value in txout_pairs)
+            for hash168 in hash168s:
+                self.hash168s[hash168].remove(hex_hash)
+            touched.update(hash168s)
+        if gone:
+            self.logger.info('{:,d} entries removed from mempool'
+                             .format(len(gone)))
+
+        # Get the raw transactions for the new hashes.  Ignore the
+        # ones the daemon no longer has (it will return None).  Put
+        # them into a dictionary of hex hash to deserialized tx.
+        hex_hashes.difference_update(self.txs)
+        raw_txs = await self.bp.daemon.getrawtransactions(hex_hashes)
+        new_txs = {hex_hash: Deserializer(raw_tx).read_tx()
+                   for hex_hash, raw_tx in zip(hex_hashes, raw_txs) if raw_tx}
+        del raw_txs, hex_hashes
+
+        # The mempool is unordered, so process all outputs first so
+        # that looking for inputs has full info.
+        parse_script = ScriptPubKey.from_script
+        coin = self.bp.coin
+        utxo_lookup = self.bp.utxo_cache.lookup
+
+        def txout_pair(txout):
+            return (parse_script(txout.pk_script, coin).hash168, txout.value)
+
+        for hex_hash, tx in new_txs.items():
+            txout_pairs = tuple(txout_pair(txout) for txout in tx.outputs)
+            self.txs[hex_hash] = [None, txout_pairs, None]
+
+        def txin_info(txin):
+            hex_hash = hash_to_str(txin.prev_hash)
+            mempool_entry = self.txs.get(hex_hash)
+            if mempool_entry:
+                return mempool_entry[1][txin.prev_idx], True
+            entry = utxo_lookup(txin.prev_hash, txin.prev_idx)
+            if entry == NO_CACHE_ENTRY:
+                # Not possible unless daemon is lying or we're corrupted?
+                self.logger.warning('no UTXO found for {} / {}'
+                                    .format(hash_to_str(txin.prev_hash),
+                                            txin.prev_idx))
+                raise MissingUTXOError
+            value, = struct.unpack('<Q', entry[-8:])
+            return (entry[:21], value), False
+
+        # Now add the inputs
+        for hex_hash, tx in new_txs.items():
+            txout_pairs = self.txs[hex_hash][1]
+            try:
+                infos = (txin_info(txin) for txin in tx.inputs)
+                txin_pairs, unconfs = zip(*infos)
+            except MissingUTXOError:
+                # If we were missing a UTXO for some reason drop this tx
+                del self.txs[hex_hash]
+                continue
+            self.txs[hex_hash] = [txin_pairs, txout_pairs, any(unconfs)]
+
+            # Update touched and self.hash168s for the new tx
+            for hash168, value in txin_pairs:
+                self.hash168s[hash168].add(hex_hash)
+                touched.add(hash168)
+            for hash168, value in txout_pairs:
+                self.hash168s[hash168].add(hex_hash)
+                touched.add(hash168)
+
+        self.logger.info('{:,d} entries in mempool for {:,d} addresses'
+                         .format(len(self.txs), len(self.hash168s)))
+
+        # Might include a None
+        return touched
+
+    def transactions(self, hash168):
+        '''Generate (hex_hash, tx_fee, unconfirmed) tuples for mempool
+        entries for the hash168.
+
+        unconfirmed is True if any txin is confirmed.
+        '''
+        for hex_hash in self.hash168s[hash168]:
+            txin_pairs, txout_pairs, unconfirmed = self.txs[hex_hash]
+            tx_fee = (sum(v for hash168, v in txin_pairs)
+                      - sum(v for hash168, v in txout_pairs))
+            yield (hex_hash, tx_fee, unconfirmed)
+
+    def value(self, hash168):
+        '''Return the unconfirmed amount in the mempool for hash168.
+
+        Can be positive or negative.
+        '''
+        value = 0
+        for tx_hash in self.hash168s[hash168]:
+            txin_pairs, txout_pairs, unconfirmed = self.txs[tx_hash]
+            value -= sum(v for h168, v in txin_pairs if h168 == hash168)
+            value += sum(v for h168, v in txout_pairs if h168 == hash168)
+        return value
+
 
 class BlockProcessor(LoggedClass):
     '''Process blocks and update the DB state to match.
@@ -153,6 +291,8 @@ class BlockProcessor(LoggedClass):
 
         self.daemon = daemon
         self.on_update = on_update
+        self.mempool = MemPool(self)
+        self.touched = set()
 
         # Meta
         self.utxo_MB = env.utxo_MB
@@ -232,27 +372,29 @@ class BlockProcessor(LoggedClass):
         Blocks are only processed in the forward direction.  The
         prefetcher only provides a non-None mempool when caught up.
         '''
-        all_touched = [set()]
-        blocks, mempool = await self.prefetcher.get_blocks()
+        blocks, mempool_hashes = await self.prefetcher.get_blocks()
+        caught_up = mempool_hashes is not None
         for block in blocks:
-            touched = self.advance_block(block)
-            if touched is None:
-                all_touched.append(await self.handle_chain_reorg())
-                mempool = None
-                break
-            all_touched.append(touched)
+            if self.advance_block(block, caught_up):
+                await self.handle_chain_reorg()
+                return
             await asyncio.sleep(0)  # Yield
 
-        if mempool is not None:
-            # Caught up to daemon height.  Flush everything as queries
-            # are performed on the DB and not in-memory.
-            self.flush(True)
-            if self.first_sync:
-                self.first_sync = False
-                self.logger.info('synced to height {:,d}'.format(self.height))
-            if self.on_update:
-                await self.on_update(self.height, set.union(*all_touched))
+        if caught_up:
+            await self.caught_up(mempool_hashes)
 
+    async def caught_up(self, mempool_hashes):
+        '''Called after each deamon poll if caught up.'''
+        # Caught up to daemon height.  Flush everything as queries
+        # are performed on the DB and not in-memory.
+        self.flush(True)
+        if self.first_sync:
+            self.first_sync = False
+            self.logger.info('synced to height {:,d}'.format(self.height))
+        if self.on_update:
+            self.touched.update(await self.mempool.update(mempool_hashes))
+            await self.on_update(self.height, self.touched)
+        self.touched = set()
 
     async def force_chain_reorg(self, to_genesis):
         try:
@@ -266,20 +408,17 @@ class BlockProcessor(LoggedClass):
         self.flush(True)
         self.logger.info('finding common height...')
 
-        touched = set()
         hashes = await self.reorg_hashes(to_genesis)
         # Reverse and convert to hex strings.
         hashes = [hash_to_str(hash) for hash in reversed(hashes)]
         for hex_hashes in chunks(hashes, 50):
             blocks = await self.daemon.raw_blocks(hex_hashes)
-            touched.update(self.backup_blocks(blocks))
+            self.backup_blocks(blocks)
 
         self.logger.info('backed up to height {:,d}'.format(self.height))
         await self.prefetcher.clear(self.height)
         self.logger.info('prefetcher reset')
 
-        return touched
-
     async def reorg_hashes(self, to_genesis):
         '''Return the list of hashes to back up beacuse of a reorg.
 
@@ -565,7 +704,7 @@ class BlockProcessor(LoggedClass):
         '''Read undo information from a file for the current height.'''
         return self.db.get(self.undo_key(height))
 
-    def advance_block(self, block):
+    def advance_block(self, block, update_touched):
         # We must update the fs_cache before calling advance_txs() as
         # the UTXO cache uses the fs_cache via get_tx_hash() to
         # resolve compressed key collisions
@@ -590,7 +729,8 @@ class BlockProcessor(LoggedClass):
             if utxo_MB >= self.utxo_MB or hist_MB >= self.hist_MB:
                 self.flush(utxo_MB >= self.utxo_MB)
 
-        return touched
+        if update_touched:
+            self.touched.update(touched)
 
     def advance_txs(self, tx_hashes, txs, touched):
         put_utxo = self.utxo_cache.put
@@ -661,9 +801,9 @@ class BlockProcessor(LoggedClass):
 
         self.logger.info('backed up to height {:,d}'.format(self.height))
 
+        self.touched.update(touched)
         flush_history = partial(self.backup_history, hash168s=touched)
         self.flush(True, flush_history=flush_history)
-        return touched
 
     def backup_txs(self, tx_hashes, txs, touched):
         # Prevout values, in order down the block (coinbase first if present)
@@ -704,9 +844,24 @@ class BlockProcessor(LoggedClass):
         assert isinstance(limit, int) and limit >= 0
         return limit
 
+    def mempool_transactions(self, hash168):
+        '''Generate (hex_hash, tx_fee, unconfirmed) tuples for mempool
+        entries for the hash168.
+
+        unconfirmed is True if any txin is confirmed.
+        '''
+        return self.mempool.transactions(hash168)
+
+    def mempool_value(self, hash168):
+        '''Return the unconfirmed amount in the mempool for hash168.
+
+        Can be positive or negative.
+        '''
+        return self.mempool.value(hash168)
+
     def get_history(self, hash168, limit=1000):
         '''Generator that returns an unpruned, sorted list of (tx_hash,
-        height) tuples of transactions that touched the address,
+        height) tuples of confirmed transactions that touched the address,
         earliest in the blockchain first.  Includes both spending and
         receiving transactions.  By default yields at most 1000 entries.
         Set limit to None to get them all.
@@ -756,7 +911,7 @@ class BlockProcessor(LoggedClass):
         hash168 = None
         if 0 <= index <= 65535:
             idx_packed = struct.pack('<H', index)
-            hash168 = self.utxo_cache.hash168(tx_hash, idx_packed)
+            hash168 = self.utxo_cache.hash168(tx_hash, idx_packed, False)
             if hash168 == NO_CACHE_ENTRY:
                 hash168 = None
         return hash168
diff --git a/server/cache.py b/server/cache.py
index 50b424c..29931f4 100644
--- a/server/cache.py
+++ b/server/cache.py
@@ -95,39 +95,41 @@ class UTXOCache(LoggedClass):
         self.cache_spends = 0
         self.db_deletes = 0
 
-    def spend(self, prev_hash, prev_idx):
-        '''Spend a UTXO and return the cache's value.
+    def lookup(self, prev_hash, prev_idx):
+        '''Given a prevout, return a pair (hash168, value).
 
-        If the UTXO is not in the cache it must be on disk.
-        '''
-        # Fast track is it's in the cache
-        pack = struct.pack
-        idx_packed = pack('<H', prev_idx)
-        value = self.cache.pop(prev_hash + idx_packed, None)
+        If the UTXO is not found, returns (None, None).'''
+        # Fast track is it being in the cache
+        idx_packed = struct.pack('<H', prev_idx)
+        value = self.cache.get(prev_hash + idx_packed, None)
         if value:
-            self.cache_spends += 1
             return value
+        return self.db_lookup(prev_hash, idx_packed, False)
 
-        # Oh well.  Find and remove it from the DB.
-        hash168 = self.hash168(prev_hash, idx_packed, True)
+    def db_lookup(self, tx_hash, idx_packed, delete=True):
+        '''Return a UTXO from the DB.  Remove it if delete is True.
+
+        Return NO_CACHE_ENTRY if it is not in the DB.'''
+        hash168 = self.hash168(tx_hash, idx_packed, delete)
         if not hash168:
             return NO_CACHE_ENTRY
 
-        self.db_deletes += 1
-
         # Read the UTXO through the cache from the disk.  We have to
         # go through the cache because compressed keys can collide.
-        key = b'u' + hash168 + prev_hash[:UTXO_TX_HASH_LEN] + idx_packed
+        key = b'u' + hash168 + tx_hash[:UTXO_TX_HASH_LEN] + idx_packed
         data = self.cache_get(key)
         if data is None:
             # Uh-oh, this should not happen...
             self.logger.error('found no UTXO for {} / {:d} key {}'
-                             .format(hash_to_str(prev_hash), prev_idx,
+                             .format(hash_to_str(tx_hash),
+                                     struct.unpack('<H', idx_packed),
                                      bytes(key).hex()))
             return NO_CACHE_ENTRY
 
         if len(data) == 12:
-            self.cache_delete(key)
+            if delete:
+                self.db_deletes += 1
+                self.cache_delete(key)
             return hash168 + data
 
         # Resolve the compressed key collison.  These should be
@@ -135,26 +137,42 @@ class UTXOCache(LoggedClass):
         assert len(data) % 12 == 0
         for n in range(0, len(data), 12):
             (tx_num, ) = struct.unpack('<I', data[n:n+4])
-            tx_hash, height = self.parent.get_tx_hash(tx_num)
-            if prev_hash == tx_hash:
-                result = hash168 + data[n: n+12]
-                data = data[:n] + data[n+12:]
-                self.cache_write(key, data)
+            this_tx_hash, height = self.parent.get_tx_hash(tx_num)
+            if tx_hash == this_tx_hash:
+                result = hash168 + data[n:n+12]
+                if delete:
+                    self.db_deletes += 1
+                    self.cache_write(key, data[:n] + data[n+12:])
                 return result
 
         raise Exception('could not resolve UTXO key collision')
 
-    def hash168(self, tx_hash, idx_packed, delete=False):
+    def spend(self, prev_hash, prev_idx):
+        '''Spend a UTXO and return the cache's value.
+
+        If the UTXO is not in the cache it must be on disk.
+        '''
+        # Fast track is it being in the cache
+        idx_packed = struct.pack('<H', prev_idx)
+        value = self.cache.pop(prev_hash + idx_packed, None)
+        if value:
+            self.cache_spends += 1
+            return value
+
+        return self.db_lookup(prev_hash, idx_packed)
+
+    def hash168(self, tx_hash, idx_packed, delete=True):
         '''Return the hash168 paid to by the given TXO.
 
-        Refers to the database.  Returns None if not found (which is
-        indicates a non-standard script).
+        Look it up in the DB and removes it if delete is True.  Return
+        None if not found.
         '''
         key = b'h' + tx_hash[:ADDR_TX_HASH_LEN] + idx_packed
         data = self.cache_get(key)
         if data is None:
-            # Assuming the DB is not corrupt, this indicates a
-            # successful spend of a non-standard script
+            # Assuming the DB is not corrupt, if delete is True this
+            # indicates a successful spend of a non-standard script
+            # as we don't currently record those
             return None
 
         if len(data) == 25:
diff --git a/server/daemon.py b/server/daemon.py
index 5fddde6..e2406d1 100644
--- a/server/daemon.py
+++ b/server/daemon.py
@@ -13,7 +13,7 @@ import json
 
 import aiohttp
 
-from lib.util import LoggedClass
+import lib.util as util
 
 
 class DaemonError(Exception):
@@ -21,7 +21,7 @@ class DaemonError(Exception):
     cannot be remedied by retrying.'''
 
 
-class Daemon(LoggedClass):
+class Daemon(util.LoggedClass):
     '''Handles connections to a daemon at the given URL.'''
 
     def __init__(self, url):
@@ -107,6 +107,19 @@ class Daemon(LoggedClass):
         '''Return the serialized raw transaction with the given hash.'''
         return await self.send_single('getrawtransaction', (hex_hash, 0))
 
+    async def getrawtransactions(self, hex_hashes):
+        '''Return the serialized raw transactions with the given hashes.
+
+        Breaks large requests up.  Yields after each sub request.'''
+        param_lists = tuple((hex_hash, 0) for hex_hash in hex_hashes)
+        raw_txs = []
+        for chunk in util.chunks(param_lists, 10000):
+            txs = await self.send_vector('getrawtransaction', chunk)
+            # Convert hex strings to bytes
+            raw_txs.append(tuple(bytes.fromhex(tx) for tx in txs))
+            await asyncio.sleep(0)
+        return sum(raw_txs, ())
+
     async def sendrawtransaction(self, params):
         '''Broadcast a transaction to the network.'''
         return await self.send_single('sendrawtransaction', params)
diff --git a/server/protocol.py b/server/protocol.py
index 23f8333..69105e2 100644
--- a/server/protocol.py
+++ b/server/protocol.py
@@ -255,9 +255,15 @@ class ElectrumX(JSONRPC):
     @classmethod
     def address_status(cls, hash168):
         '''Returns status as 32 bytes.'''
+        # Note history is ordered and mempool unordered in electrum-server
+        # For mempool, height is -1 if unconfirmed txins, otherwise 0
         history = cls.BLOCK_PROCESSOR.get_history(hash168)
+        mempool = cls.BLOCK_PROCESSOR.mempool_transactions(hash168)
+
         status = ''.join('{}:{:d}:'.format(hash_to_str(tx_hash), height)
                          for tx_hash, height in history)
+        status += ''.join('{}:{:d}:'.format(hex_hash, -unconfirmed)
+                          for hex_hash, tx_fee, unconfirmed in mempool)
         if status:
             return sha256(status.encode()).hex()
         return None
@@ -297,11 +303,16 @@ class ElectrumX(JSONRPC):
 
     @classmethod
     def get_history(cls, hash168):
+        # Note history is ordered and mempool unordered in electrum-server
+        # For mempool, height is -1 if unconfirmed txins, otherwise 0
         history = cls.BLOCK_PROCESSOR.get_history(hash168, limit=None)
-        return [
-            {'tx_hash': hash_to_str(tx_hash), 'height': height}
-            for tx_hash, height in history
-        ]
+        mempool = cls.BLOCK_PROCESSOR.mempool_transactions(hash168)
+
+        conf = tuple({'tx_hash': hash_to_str(tx_hash), 'height': height}
+                       for tx_hash, height in history)
+        unconf = tuple({'tx_hash': tx_hash, 'height': -unconfirmed, 'fee': fee}
+                       for tx_hash, fee, unconfirmed in mempool)
+        return conf + unconf
 
     @classmethod
     def get_chunk(cls, index):
@@ -315,7 +326,7 @@ class ElectrumX(JSONRPC):
     @classmethod
     def get_balance(cls, hash168):
         confirmed = cls.BLOCK_PROCESSOR.get_balance(hash168)
-        unconfirmed = -1  # FIXME
+        unconfirmed = cls.BLOCK_PROCESSOR.mempool_value(hash168)
         return {'confirmed': confirmed, 'unconfirmed': unconfirmed}
 
     @classmethod
diff --git a/server/version.py b/server/version.py
index ef70379..8acc7f9 100644
--- a/server/version.py
+++ b/server/version.py
@@ -1 +1 @@
-VERSION = "ElectrumX 0.05"
+VERSION = "ElectrumX 0.06"