diff --git a/README.rst b/README.rst
index 0808ea0..00bc6c5 100644
--- a/README.rst
+++ b/README.rst
@@ -1,13 +1,127 @@
-===============================
-guarneri
-===============================
+==========================
+Guarneri Instrument Maker
+==========================
 
 
 .. image:: https://img.shields.io/pypi/v/guarneri.svg
         :target: https://pypi.python.org/pypi/guarneri
 
 
-guarneri
+A package for creating Ophyd and Ophyd-async devices from configuration
+files.
+
+Instead of instantiating devices directly in python, Guarneri reads a
+configuration file and creates/connects the devices for you. This
+provides the following benefits:
+
+1) Beamline configuration is in a human-readable configuration file (e.g. TOML).
+2) Other tools can modify the configuration file if needed.
+3) Devices can be connected in parallel.
+4) Missing devices are handled gracefully.
+5) Devices can be simulated/mocked by changing single value in the config file.
+
+
+Usage
+-----
+
+Let's say you have some ophyd and ophyd-async devices defined (with
+type hints) in a file called ``devices.py``.
+
+.. code-block:: python
+   :caption: devices.py
+
+    from ophyd_async.epics import epics_signal_rw
+    from ophyd_async.core import AsyncDevice
+    from ophyd import Device, Component
+    
+    from guarneri import Instrument
+    
+    class MyDevice(Device):
+        description = Component(".DESC")
+    
+    
+    class MyAsyncDevice(AsyncDevice):
+        def __init__(self, prefix: str, name: str = ""):
+            self.description = epics_signal_rw(str, f"{prefix}.DESC")
+    	super().__init__(name=name)
+
+
+    def area_detector_factory(hdf: bool=True):
+        # Create devices here using the arguments
+        area_detector = ...
+	return area_detector
+
+
+Instead of instantiating these in a python startup script, Guarneri
+will let you **create devices from a TOML configuration file**. First
+we create a TOML file with the necessary parameters, these map
+directly onto the arguments of the device's ``__init__()``, or the
+arguments of a factory that returns a device.
+
+
+.. code-block:: toml
+   :caption: instrument.toml
+
+    [my_device.device1]
+    prefix = '255id:'
+    
+    [async_device.device3]
+    prefix = '255id:'
+
+    [area_detector.sim_det]
+    hdf = true
+
+
+Then in your beamline startup code, create a Guarneri instrument and
+load the config files.
+
+.. code-block:: python
+   :caption: instrument.py
+
+    from io import StringIO
+		
+    from devices import MyDevice, MyAsyncDevice, area_detector_factory
+    
+    # Prepare the instrument device
+    instrument = Instrument({
+        "my_device": MyDevice,
+	"async_device": MyAsyncDevice,
+	"area_detector": area_detector_factory,
+    })
+
+    # Create the devices from the TOML configuration file
+    instrument.load_config_files("instrument.toml")
+    # Optionally connect all the devices
+    await instrument.connect_devices()
+
+    # Now use the devices for science!
+    instrument.devices['device_1'].description.get()
+
+
+The first argument to :py:meth:`guarneri.Instrument.__init__()` is a
+mapping of TOML section names to device classes. Guarneri then
+introspects the device or factory to decide which TOML keys and types
+are valid. In the above example, the heading ``[my_device.device1]``
+will create an instance of ``MyDevice()`` with the name ``"device1"``.
+
+
+What About Happi?
+-----------------
+
+Happi has a similar goal to Guarneri, but with a different
+scope. While Happi is meant for facility-level configuration (e.g.
+LCLS), Guarneri is aimed at individual beamlines at a synchrotron.
+Happi uses :py:class:`HappiItem` classes with :py:class:`ItemInfo`
+objects to describe the devices definitions, while Guarneri uses the
+device classes themselves. Happi provides a python client for adding
+and modifying the devices, while Guarneri uses human-readable
+configuration files.
+
+**Which one is better?** Depends on what you're trying to do. If you
+want a **flexible and scalable** system the **shares devices across a
+facility**, try Happi. If you want a way to **get devices running
+quickly** on your beamline before users show up, try Guarneri.
+
 
 Documentation
 -------------