diff --git a/cluster/aws/options.md b/cluster/aws/options.md
new file mode 100644
index 00000000000..fcc5e01cdce
--- /dev/null
+++ b/cluster/aws/options.md
@@ -0,0 +1,24 @@
+# AWS specific configuration options
+
+These options can be set as environment variables to customize how your cluster is created.  Only options
+specific to AWS are documented in this file, for cross-provider options see TODO:global-options-doc.
+
+This is a work-in-progress; not all options are documented yet!
+
+## AWS_IMAGE
+
+The AMI to use.  If not specified, the image will be selected based on the AWS region.
+
+## AWS_S3_BUCKET, AWS_S3_REGION
+
+The bucket name to use, and the region where the bucket should be created, or where the bucket is located if it exists already.
+
+If not specified, defaults to AWS_S3_REGION us-east-1, because buckets are globally named and you probably
+want to share a bucket across all regions; us-east-1 is a sensible (relatively arbitrary) default.
+
+AWS_S3_BUCKET will default to a uniquely generated name, so you won't collide with other kubernetes users.
+(Currently this uses the hash of your AWS Access key to produce a per-user unique value).
+
+It is not a bad idea to set AWS_S3_BUCKET to something more human friendly.
+
+AWS_S3_REGION is useful for people that want to control their data location, because of regulatory restrictions for example.