diff --git a/nilmdb/client/numpyclient.py b/nilmdb/client/numpyclient.py
new file mode 100644
index 0000000..5b02a97
--- /dev/null
+++ b/nilmdb/client/numpyclient.py
@@ -0,0 +1,77 @@
+# -*- coding: utf-8 -*-
+
+"""Provide a NumpyClient class that is based on normal Client, but has
+additional methods for extracting and inserting data via Numpy arrays."""
+
+import nilmdb.utils
+import nilmdb.client.client
+import nilmdb.client.httpclient
+from nilmdb.client.errors import ClientError
+
+import contextlib
+from nilmdb.utils.time import timestamp_to_string, string_to_timestamp
+
+import numpy
+import cStringIO
+
+def layout_to_dtype(layout):
+    ltype = layout.split('_')[0]
+    lcount = int(layout.split('_')[1])
+    if ltype.startswith('int'):
+        atype = '<i' + str(int(ltype[3:]) / 8)
+    elif ltype.startswith('uint'):
+        atype = '<u' + str(int(ltype[4:]) / 8)
+    elif ltype.startswith('float'):
+        atype = '<f' + str(int(ltype[5:]) / 8)
+    else:
+        raise ValueError("bad layout")
+    return numpy.dtype([('timestamp', '<i8'), ('data', atype, lcount)])
+
+class NumpyClient(nilmdb.client.client.Client):
+    """Subclass of nilmdb.client.Client that adds additional methods for
+    extracting and inserting data via Numpy arrays."""
+
+    def stream_extract_numpy(self, path, start = None, end = None,
+                             layout = None, maxrows = 100000,
+                             structured = False):
+        """
+        Extract data from a stream.  Returns a generator that yields
+        Numpy arrays of up to 'maxrows' of data each.
+
+        If 'layout' is None, it is read using stream_info.
+
+        If 'structured' is False, all data is converted to float64
+        and returned in a flat 2D array.  Otherwise, data is returned
+        as a structured dtype in a 1D array.
+        """
+        if layout is None:
+            streams = self.stream_list(path)
+            if len(streams) != 1:
+                raise ClientError("can't get layout for path: " + path)
+            layout = streams[0][1]
+        dtype = layout_to_dtype(layout)
+
+        def to_numpy(data):
+            a = numpy.fromstring(data, dtype)
+            if structured:
+                return a
+            return numpy.c_[a['timestamp'], a['data']]
+
+        chunks = []
+        total_len = 0
+        maxsize = dtype.itemsize * maxrows
+        for data in self.stream_extract(path, start, end, binary = True):
+            # Add this block of binary data
+            chunks.append(data)
+            total_len += len(data)
+
+            # See if we have enough to make the requested Numpy array
+            while total_len >= maxsize:
+                assembled = "".join(chunks)
+                total_len -= maxsize
+                chunks = [ assembled[maxsize:] ]
+                block = assembled[:maxsize]
+                yield to_numpy(block)
+
+        if total_len:
+            yield to_numpy("".join(chunks))
diff --git a/tests/test.order b/tests/test.order
index 59e528e..0b3f06e 100644
--- a/tests/test.order
+++ b/tests/test.order
@@ -12,6 +12,7 @@ test_interval.py
 test_bulkdata.py
 test_nilmdb.py
 test_client.py
+test_numpyclient.py
 test_cmdline.py
 
 test_*.py
diff --git a/tests/test_numpyclient.py b/tests/test_numpyclient.py
new file mode 100644
index 0000000..915762d
--- /dev/null
+++ b/tests/test_numpyclient.py
@@ -0,0 +1,108 @@
+# -*- coding: utf-8 -*-
+
+import nilmdb.server
+import nilmdb.client
+import nilmdb.client.numpyclient
+
+from nilmdb.utils.printf import *
+from nilmdb.utils import timestamper
+from nilmdb.client import ClientError, ServerError
+from nilmdb.utils import datetime_tz
+
+from nose.plugins.skip import SkipTest
+from nose.tools import *
+from nose.tools import assert_raises
+import itertools
+import distutils.version
+
+from testutil.helpers import *
+
+import numpy as np
+
+testdb = "tests/numpyclient-testdb"
+testurl = "http://localhost:32180/"
+
+def setup_module():
+    global test_server, test_db
+    # Clear out DB
+    recursive_unlink(testdb)
+
+    # Start web app on a custom port
+    test_db = nilmdb.utils.serializer_proxy(nilmdb.server.NilmDB)(testdb)
+    test_server = nilmdb.server.Server(test_db, host = "127.0.0.1",
+                                       port = 32180, stoppable = False,
+                                       fast_shutdown = True,
+                                       force_traceback = True)
+    test_server.start(blocking = False)
+
+def teardown_module():
+    global test_server, test_db
+    # Close web app
+    test_server.stop()
+    test_db.close()
+
+class TestNumpyClient(object):
+
+    def test_numpyclient_01_basic(self):
+        # Test basic connection
+        client = nilmdb.client.numpyclient.NumpyClient(url = testurl)
+        version = client.version()
+        eq_(distutils.version.LooseVersion(version),
+            distutils.version.LooseVersion(test_server.version))
+
+        # Verify subclassing
+        assert(isinstance(client, nilmdb.client.Client))
+
+        # Layouts
+        for layout in "int8_t", "something_8", "integer_1":
+            with assert_raises(ValueError):
+                for x in client.stream_extract_numpy("/foo", layout=layout):
+                    pass
+        for layout in "int8_1", "uint8_30", "int16_20", "float64_100":
+            with assert_raises(ClientError) as e:
+                for x in client.stream_extract_numpy("/foo", layout=layout):
+                    pass
+            in_("No such stream", str(e.exception))
+
+        with assert_raises(ClientError) as e:
+            for x in client.stream_extract_numpy("/foo"):
+                pass
+        in_("can't get layout for path", str(e.exception))
+
+        client.close()
+
+    def test_numpyclient_02_extract(self):
+        client = nilmdb.client.numpyclient.NumpyClient(url = testurl)
+
+        # Insert some data as text
+        client.stream_create("/newton/prep", "float32_8")
+        testfile = "tests/data/prep-20120323T1000"
+        start = nilmdb.utils.time.parse_time("20120323T1000")
+        rate = 120
+        data = timestamper.TimestamperRate(testfile, start, rate)
+        result = client.stream_insert("/newton/prep", data,
+                                      start, start + 119999777)
+
+        # Extract Numpy arrays
+        array = None
+        pieces = 0
+        for chunk in client.stream_extract_numpy("/newton/prep", maxrows=1000):
+            pieces += 1
+            if array is not None:
+                array = np.vstack((array, chunk))
+            else:
+                array = chunk
+        eq_(array.shape, (14400, 9))
+        eq_(pieces, 15)
+
+        # Try structured
+        s = list(client.stream_extract_numpy("/newton/prep", structured = True))
+        assert(np.array_equal(np.c_[s[0]['timestamp'], s[0]['data']], array))
+
+        # Compare.  Will be close but not exact because the conversion
+        # to and from ASCII was lossy.
+        data = timestamper.TimestamperRate(testfile, start, rate)
+        actual = np.fromstring(" ".join(data), sep=' ').reshape(14400, 9)
+        assert(np.allclose(array, actual))
+
+        client.close()