diff --git a/docs/feature_specs/kafka/extractor.md b/docs/feature_specs/kafka/extractor.md
new file mode 100644
index 000000000..2d1009e6c
--- /dev/null
+++ b/docs/feature_specs/kafka/extractor.md
@@ -0,0 +1,20 @@
+# Kafka - data extractor
+
+The data extractor is responsible for loading data from Kafka.  In order to do
+so, it needs to know the URI of the Kafka leader broker.  Once the extractor
+connects to Kafka, it starts importing data.
+
+Data extractor depends on [cppkafka](https://github.com/mfontanini/cppkafka)
+which makes message consumption just a few API calls, as seen
+[here](https://github.com/mfontanini/cppkafka/wiki/Consuming-messages).
+
+There are also other metadata that can be passed to data extractor that are
+defined with our [extension](opencypher.md) of openCypher.
+
+A full list of configurable metadata can be found
+[here](https://github.com/edenhill/librdkafka/blob/master/CONFIGURATION.md).
+
+Memgraph supports customizing the following:
+
+  * `metadata.broker.list` which is a required parameter, set by `KAFKA 'URI'`
+  * `queue.buffering.max.ms` set by `BATCH INTERVAL`
diff --git a/docs/feature_specs/kafka/opencypher.md b/docs/feature_specs/kafka/opencypher.md
new file mode 100644
index 000000000..22f9dd2a1
--- /dev/null
+++ b/docs/feature_specs/kafka/opencypher.md
@@ -0,0 +1,52 @@
+# Kafka - openCypher clause
+
+One must be able to specify the following when importing data from Kafka:
+* Kafka URI
+* Transform [script](transform.md) URI
+
+
+Kafka endpoint is the URI of the leader broker and it is required for data
+[extractor](extractor.md).
+
+Minimum required syntax looks like:
+```opencypher
+CREATE STREAM kafka_stream AS LOAD DATA KAFKA '127.0.0.1/topic' WITH TRANSFORM
+'127.0.0.1/transform.py';
+```
+
+The `CREATE STREAM` clause happens in a transaction.
+
+The full openCypher clause for creating a stream is:
+```opencypher
+CREATE STREAM stream_name AS
+  LOAD DATA KAFKA 'URI'
+  WITH TRANSFORM 'URI'
+  [BATCH INTERVAL milliseconds]
+```
+
+The `WITH TRANSFORM` parameter should contain a URI of the transform script.
+
+The `BATCH_INTERVAL` parameter defines the time interval in milliseconds
+that defines the time between two successive stream importing operations.
+
+The `DROP` clause deletes a stream:
+```opencypher
+DROP STREAM stream_name;
+```
+
+The `SHOW` clause enables you to see all configured streams:
+```opencypher
+SHOW STREAMS;
+```
+
+You can also start/stop streams with the `START` and `STOP` clauses:
+```opencypher
+START STREAM stream_name [LIMIT count BATCHES];
+STOP STREAM stream_name;
+```
+
+There are also convenience clauses to start and stop all streams:
+```opencypher
+START ALL STREAMS;
+STOP ALL STREAMS;
+```
diff --git a/docs/feature_specs/kafka/transform.md b/docs/feature_specs/kafka/transform.md
new file mode 100644
index 000000000..92b870c89
--- /dev/null
+++ b/docs/feature_specs/kafka/transform.md
@@ -0,0 +1,52 @@
+# Kafka - data transform
+
+The transform script is a user defined script written in Python.  The script
+should be aware of the data format in the Kafka message.
+
+Each Kafka message is byte length encoded, which means that the first eight
+bytes of each message contain the length of the message.
+
+More on the message format can be seen
+[here](https://cwiki.apache.org/confluence/display/KAFKA/A+Guide+To+The+Kafka+Protocol#AGuideToTheKafkaProtocol-Messagesets).
+
+
+The script will be embedded in our C++ codebase using pythons
+[embedding](https://docs.python.org/3.5/extending/embedding.html) feature.
+
+A sample code for a streaming transform script could look like this:
+
+```python
+import struct
+import sys
+
+def get_records():
+    while True:
+        message_len = sys.stdin.read(8)
+        if len(message_len) == 8:
+            message_len = struct.unpack("L", message_len)[0]
+            record = sys.stdin.read(message_len)
+            yield record
+        else:
+            assert len(message_len) == 0, message_len
+            return
+
+def create_vertex(fields):
+  return "CREATE (n:Node {{id: {}}})".format(fields[1])
+
+
+def create_edge(fields):
+  return "MATCH (n:Node {{id: {}}}) "\
+         "MATCH ((m:Node {{id : {}}})) "\
+         "CREATE (n)-[e:Edge{{value: {}}}]->(m) "\
+         .format(fields[1], fields[2], fields[3])
+
+for record in get_records():
+  fields = record.split("\t")
+  if fields[0] == "v":
+    return create_vertex(fields):
+  else:
+    return create_edge(fields)
+```
+
+The script should output openCypher query strings based on the type of the
+records.