diff --git a/docs/index.rst b/docs/index.rst
index 1a5915cc5dd6..28d18bff8f00 100644
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -69,6 +69,13 @@
   dns-resource_record_set
   dns-changes
 
+.. toctree::
+  :maxdepth: 0
+  :hidden:
+  :caption: Cloud Search
+
+  search-usage
+
 .. toctree::
   :maxdepth: 0
   :hidden:
diff --git a/docs/search-usage.rst b/docs/search-usage.rst
new file mode 100644
index 000000000000..9d2ee0c05152
--- /dev/null
+++ b/docs/search-usage.rst
@@ -0,0 +1,199 @@
+Using the API
+=============
+
+Overview
+~~~~~~~~
+
+Cloud Search allows an application to quickly perform full-text and
+geospatial searches without having to spin up instances
+and without the hassle of managing and maintaining a search service.
+
+Cloud Search provides a model for indexing documents containing structured data,
+with documents and indexes saved to a separate persistent store optimized
+for search operations.
+
+The API supports full text matching on string fields and allows indexing
+any number of documents in any number of indexes.
+
+Client
+------
+
+:class:`Client <gcloud.search.client.Client>` objects provide a means to
+configure your Cloud Search applications.  Eash instance holds both a
+``project`` and an authenticated connection to the Cloud Search service.
+
+For an overview of authentication in ``gcloud-python``, see :doc:`gcloud-auth`.
+
+Assuming your environment is set up as described in that document,
+create an instance of :class:`Client <gcloud.search.client.Client>`.
+
+.. doctest::
+
+   >>> from gcloud import search
+   >>> client = search.Client()
+
+Indexes
+~~~~~~~
+
+Indexes are searchable collections of documents.
+
+List all indexes in the client's project:
+
+.. doctest::
+
+   >>> indexes = client.list_indexes()  # API call
+   >>> for index in indexes:
+   ...     print(index.name)
+   ...     field_names = ', '.join([field.name for field in index.fields])
+   ...     print('- %s' % field_names)
+   index-name
+   - field-1, field-2
+   another-index-name
+   - field-3
+
+Create a new index:
+
+.. doctest::
+
+   >>> new_index = client.index('new-index-name')
+
+.. note::
+
+   Indexes cannot be created, updated, or deleted directly on the server:
+   they are derived from the documents which are created "within" them.
+
+Documents
+~~~~~~~~~
+
+Create a document instance, which is not yet added to its index on
+the server:
+
+.. doctest::
+
+   >>> index = client.index('index-id')
+   >>> document = index.document('document-1')
+   >>> document.exists()  # API call
+   False
+   >>> document.rank
+   None
+
+Add one or more fields to the document:
+
+.. doctest::
+
+   >>> field = document.Field('fieldname')
+   >>> field.add_value('string')
+
+Save the document into the index:
+
+.. doctest::
+
+   >>> document.create()  # API call
+   >>> document.exists()  # API call
+   True
+   >>> document.rank      # set by the server
+   1443648166
+
+List all documents in an index:
+
+.. doctest::
+
+   >>> documents = index.list_documents()  # API call
+   >>> [document.id for document in documents]
+   ['document-1']
+
+Delete a document from its index:
+
+.. doctest::
+
+   >>> document = index.document('to-be-deleted')
+   >>> document.exists()  # API call
+   True
+   >>> document.delete()  # API call
+   >>> document.exists()  # API clal
+   False
+
+.. note::
+
+   To update a document in place after manipulating its fields or rank, just
+   recreate it:  E.g.:
+
+   .. doctest::
+
+      >>> document = index.document('document-id')
+      >>> document.exists()  # API call
+      True
+      >>> document.rank = 12345
+      >>> field = document.field('field-name')
+      >>> field.add_value('christina aguilera')
+      >>> document.create()  # API call
+
+Fields
+~~~~~~
+
+Fields belong to documents and are the data that actually gets searched.
+
+Each field can have multiple values, which can be of the following types:
+
+- String (Python2 :class:`unicode`, Python3 :class:`str`)
+- Number (Python :class:`int` or :class:`float`)
+- Timestamp (Python :class:`datetime.datetime`)
+- Geovalue (Python tuple, (:class:`float`, :class:`float`))
+
+String values can be tokenized using one of three different types of
+tokenization, which can be passed when the value is added:
+
+- **Atom** (``atom``) means "don't tokenize this string", treat it as one
+  thing to compare against.
+
+- **Text** (``text``) means "treat this string as normal text" and split words
+  apart to be compared against.
+
+- **HTML** (``html``) means "treat this string as HTML", understanding the
+  tags, and treating the rest of the content like Text.
+
+.. doctest::
+
+   >>> from gcloud import search
+   >>> client = search.Client()
+   >>> index = client.index('index-id')
+   >>> document = index.document('document-id')
+   >>> field = document.field('field-name')
+   >>> field.add_value('britney spears', tokenization='atom')
+   >>> field.add_value(''<h1>Britney Spears</h1>', tokenization='html')
+
+Searching
+~~~~~~~~~
+
+After populating an index with documents, search through them by
+issuing a search query:
+
+.. doctest::
+
+   >>> from gcloud import search
+   >>> client = search.Client()
+   >>> index = client.index('index-id')
+   >>> query = client.query('britney spears')
+   >>> matching_documents = index.search(query)  # API call
+   >>> for document in matching_documents:
+   ...     print(document.id)
+   ['document-id']
+
+By default, all queries are sorted by the ``rank`` value set when the
+document was created.  See:
+https://cloud.google.com/search/reference/rest/v1/projects/indexes/documents#resource_representation.google.cloudsearch.v1.Document.rank
+
+To sort differently, use the ``order_by`` parameter:
+
+.. doctest::
+
+   >>> ordered = client.query('britney spears', order_by=['field1', '-field2'])
+
+Note that the ``-`` character before ``field2`` means that this query will
+be sorted ascending by ``field1`` and then descending by ``field2``.
+
+To limit the fields to be returned in the match, use the ``fields`` paramater:
+
+.. doctest::
+
+   >>> projected = client.query('britney spears', fields=['field1', 'field2'])