Liking cljdoc? Tell your friends :D


aws-api is a Clojure library which provides programmatic access to AWS services from your Clojure program.


AWS APIs are data-oriented in both the "send data, get data back" sense, and the fact that all of the operations and data structures for every service are, themselves, described in data which can be used to generate mechanical transformations from application data to wire data and back. This is exactly what we want from a Clojure API.

Using the AWS Java SDK directly via interop requires knowledge of OO hierarchies of what are basically data classes, and while the existing Clojure wrappers hide much of this from you, they don't hide it from your process.

aws-api is an idiomatic, data-oriented Clojure library for invoking AWS APIs. While the library offers some helper and documentation functions you'll use at development time, the only functions you ever need at runtime are client, which creates a client for a given service and invoke, which invokes an operation on the service. invoke takes a map and returns a map, and works the same way for every operation on every service.


AWS APIs are described in data which specifies operations, inputs, and outputs. aws-api uses the same data descriptions to expose a data-oriented interface, using service descriptions, documentation, and specs which are generated from the source descriptions.

Each AWS SDK has its own copy of the data descriptions in their github repos. We use aws-sdk-js as the source for these, and release individual artifacts for each api. The api descriptors include the AWS api-version in their filenames (and in their data). For example you'll see both of the following files listed:


Whenever we release, we look for the descriptor with the most recent API version. If aws-sdk-js-v2.351.0 contains an update to dynamodb-2012-08-10.normal.json, or a new dynamodb descriptor with a more recent api-version, we'll make a release whose version number includes the 2.351.0 from the version of aws-sdk-js.

We also include the revision of our generator in the version. For example, indicates revision 653 of the generator, and tag v2.351.0 of aws-sdk-js.

  • See Versioning for more about how we version releases.
  • See latest releases for a list of the latest releases of api, endpoints, and all supported services.



To use aws-api in your application, you depend on, and the service(s) of your choice, e.g.

To use, for example, the s3 api, add the following to deps.edn

{:deps {       {:mvn/version "0.8.515"} {:mvn/version ""}        {:mvn/version "814.2.978.0"}}}
  • See latest releases for a list of the latest releases of api, endpoints, and all supported services.


Fire up a repl using that deps.edn, and then you can do things like this:

(require '[ :as aws])

Create a client:

(def s3 (aws/client {:api :s3}))

Ask what ops your client can perform:

(aws/ops s3)

Look up docs for an operation:

(aws/doc s3 :CreateBucket)

Tell the client to let you know when you get the args wrong:

(aws/validate-requests s3 true)

Do stuff:

(aws/invoke s3 {:op :ListBuckets})
;; http-request and http-response are in the metadata
(meta *1)

;; create a bucket in the same region as the client
(aws/invoke s3 {:op :CreateBucket :request {:Bucket "my-unique-bucket-name"}})

;; create a bucket in a region other than us-east-1
(aws/invoke s3 {:op :CreateBucket :request {:Bucket "my-unique-bucket-name-in-us-west-1"
                                            {:LocationConstraint "us-west-1"}}})

;; NOTE: be sure to create a client with region "us-west-1" when accessing that
;; bucket.

(aws/invoke s3 {:op :ListBuckets})

See the examples directory for more examples.

Responses, success, failure

Barring client side exceptions, every operation on every service will return a map. If the operation is successful, the map will be in the shape described by (-> client aws/ops op :response). If AWS indicates failure with an HTTP status code >= 400, the map will include a :cognitect.anomalies/category key, so you can check for the absence/presence of that key to determine success/failure.


The aws-api client implicitly looks up credentials the same way the java SDK does.

To provide credentials explicitly, you pass an implementation of to the client constructor fn, .e.g

(require '[ :as aws])
(def kms (aws/client {:api :kms :credentials-provider my-custom-provider}))

If you're supplying a known access-key/secret pair, you can use the basic-credentials-provider helper fn:

(require '[ :as aws]
         '[ :as credentials])

(def kms (aws/client {:api                  :kms
                      :credentials-provider (credentials/basic-credentials-provider
                                             {:access-key-id     "ABC"
                                              :secret-access-key "XYZ"})}))

See the assume role example for a more involved example using AWS STS.

Region lookup

The aws-api client looks up the region the same way the java SDK does, with an additional check for a System property named "aws.region" after it checks for the AWS_REGION environment variable and before it checks your aws configuration.

Endpoint Override

Most of the time you can create a client and it figures out the correct endpoint for you, but there are exceptions. You may want to use a proxy server or connect to a local dynamodb, or perhaps you've found a bug that you could work around if you could supply the correct endpoint. All of this can be accomplished by supplying an :endpoint-override map to the client constructor:

(def ddb (aws/client {:api :dynamodb
                      :endpoint-override {:protocol :http
                                          :hostname "localhost"
                                          :port     8000}}))


The :PostToConnection operation on the apigatewaymanagementapi client requires that you specify the API endpoint as follows:

(def client (aws/client {:api :apigatewaymanagementapi
                         :endpoint-override {:hostname "{hostname}"
                                             :path "/{stage}/@connections/"}}))

Replace {hostname} and {stage} with the hostname and the stage of the connection to which you're posting (see

The client will append the :ConnectionId in the :request map to the :path in the :endpoint-override map.


NOTE: the behavior of and changed as of release 0.8.430. See Upgrade Notes for more information.

The aws-api client uses an http-client to send requests to AWS, including any operations you invoke and fetching the region and credentials when you're running in EC2 or ECS. By default, each aws-api client uses a single, shared http-client, whose resources are managed by aws-api.


This library is open source, developed internally by Cognitect. Issues can be filed using GitHub issues for this project. Because aws-api is incorporated into products and client projects, we prefer to do development internally and are not accepting pull requests or patches.


aws-api was extracted from an internal project at Cognitect, and some contributors are missing from the commit log. Here are all the folks from Cognitect who either committed code directly, or contributed significantly to research and design:

Timothy Baldridge
David Chelimsky
Benoît Fleury
Stuart Halloway
Rich Hickey
George Kierstein
Carin Meier
Alex Miller
Michael Nygard
Ghadi Shayban
Joseph Smith
Marshall Thompson



nodename nor servname provided, or not known

This indicates that the configured endpoint is incorrect for the service/op you are trying to perform.

Remedy: check AWS Regions and Endpoints the proper endpoint and use the :endpoint-override key when creating a client, e.g.

(def s3-control {:api :s3control})
(aws/client {:api :s3control
             :endpoint-override {:hostname (str my-account-id "")}})

No known endpoint.

This indicates that the data in the lib (which is derived from endpoints.json) does not support the :api/:region combination you are trying to access.

Remedy: check AWS Regions and Endpoints, and supply the correct endpoint as described in nodename nor servname provided, or not known, above.

Ops limit reached

The underlying http-client has a :pending-ops-limit configuration which, when reached, results in an exception with the message "Ops limit reached". As of this writing, aws-api does not provide access to the http-client's configuration. Programs that encounter "Ops limit reached" can avoid it by creating separate http-clients for each aws-client. You may wish to explicitly stop ( these aws-clients when the are not longer in use to conserve resources.

S3 Issues

"Invalid 'Location' header: null"

This indicates that you are trying to access an S3 resource (bucket or object) that resides in a different region from the client's region.

Remedy: create a new s3 client in the same region you are trying to access.

Copyright and License

Copyright © 2015 Cognitect

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Can you improve this documentation? These fine people already did:
David Chelimsky, Fogus & Alex Miller
Edit on GitHub

cljdoc is a website building & hosting documentation for Clojure/Script libraries

× close