MQTT is a lightweight and widely adopted messaging protocol that is designed for constrained devices. AWS IoT support for MQTT is based on the MQTT v3.1.1 specification, with some differences. For information about how AWS IoT differs from the MQTT v3.1.1 specification, see AWS IoT differences from MQTT version 3.1.1 specification.

AWS IoT Core supports device connections that use the MQTT protocol and MQTT over WSS protocol. The AWS IoT Device SDKs support both protocols and are the recommended ways to connect devices to AWS IoT. The AWS IoT Device SDKs support the functions necessary for devices and clients to connect to and access AWS IoT Core services and they support the authentication protocols that the AWS IoT services require. For information about how to connect to AWS IoT using the AWS Device SDKs and links to examples of AWS IoT in the supported languages, see Connecting with MQTT using the AWS IoT Device SDKs. For more information about authentication methods and the port mappings for MQTT messages, see Protocols, port mappings, and authentication.

While we recommend using the AWS IoT Device SDKs to connect to AWS IoT, they are not required. If you do not use the AWS IoT Device SDKs, however, you must provide the necessary connection and communication security. Clients must send the Server Name Indication (SNI) TLS extension in the connection request. Connection attempts that don't include the SNI are refused. For more information, see Transport Security in AWS IoT. Clients that use IAM users and AWS credentials to authenticate clients must provide the correct Signature Version 4 authentication.

Connecting with MQTT using the AWS IoT Device SDKs

This section contains links to the AWS IoT Device SDKs and to the source code of sample programs that illustrate how to connect a device to AWS IoT. The sample apps linked here show how to connect to AWS IoT using the MQTT protocol and MQTT over WSS.


Using the AWS IoT C++ Device SDK to connect devices


Using the AWS IoT Device SDK for Python to connect devices


Using the AWS IoT Device SDK for JavaScript to connect devices


Using the AWS IoT Device SDK for Java to connect devices

Embedded C

Using the AWS IoT Device SDK for Embedded C to connect devices


This SDK is intended for use by experienced embedded-software developers.

MQTT Quality of Service (QoS) options

AWS IoT and the AWS IoT Device SDKs support the MQTT Quality of Service (QoS) levels 0 and 1. The MQTT protocol defines a third level of QoS, level 2, but AWS IoT does not support it. Only the MQTT protocol supports the QoS feature. HTTPS does not support QoS.

This table describes how each QoS level affects messages published to and by the message broker.

With a QoS level of...

The message is...


QoS level 0

Sent zero or more times

This level should be used for messages that are sent over reliable communication links or that can be missed without a problem.

QoS level 1

Sent at least one time, and then repeatedly until a PUBACK response is received

The message is not considered complete until the sender receives a PUBACK response to indicate successful delivery.

Using MQTT persistent sessions

Persistent sessions store a client’s subscriptions and messages, with a quality of service (QoS) of 1, that have not been acknowledged by the client. When a disconnected device reconnects to a persistent session, the session resumes, its subscriptions are reinstated, and subscribed messages received prior to the reconnection and that have not been acknowledged by the client are sent to the client.

Creating a persistent session

You create an MQTT persistent session by sending a CONNECT message and setting the cleanSession flag to 0. If no session exists for the client sending the CONNECT message, a new persistent session is created. If a session already exists for the client, the client resumes the existing session.

Operations during a persistent session

Clients use the sessionPresent attribute in the connection acknowledged (CONNACK) message to determine if a persistent session is present. If sessionPresent is 1, a persistent session is present and any stored messages for the client are delivered to the client immediately after the client receives the CONNACK, as described in Message traffic after reconnection to a persistent session. If sessionPresent is 1, there is no need for the client to resubscribe. However, if sessionPresent is 0, no persistent session is present and the client must resubscribe to its topic filters.

After the client joins a persistent session, it can publish messages and subscribe to topic filters without any additional flags on each operation.

Message traffic after reconnection to a persistent session

A persistent session represents an ongoing connection between a client and an MQTT message broker. When a client connects to the message broker using a persistent session, the message broker saves all subscriptions that the client makes during the connection. When the client disconnects, the message broker stores unacknowledged QoS 1 messages and new QoS 1 messages published to topics to which the client is subscribed. When the client reconnects to its persistent session, all subscriptions are reinstated and all stored messages are sent to the client at a maximum rate of 10 messages per second.

After reconnection, the stored messages are sent to the client, at a rate that is limited to 10 stored messages per second, along with any current message traffic until the Publish requests per second per connection limit is reached. Because the delivery rate of the stored messages is limited it will take several seconds to deliver all stored messages if a session has more than 10 stored messages to deliver after reconnection.

Ending a persistent session

The following conditions describe how persistent sessions can end.

  • When the persistent session expiration time elapses. The persistent session expiration timer starts when the message broker detects that a client has disconnected, either by the client disconnecting or the connection timing out.

  • When the client sends a CONNECT message that sets the cleanSession flag to 1.


The stored messages waiting to be sent to the client when a session ends are discarded; however, they are still billed at the standard messaging rate, even though they could not be sent. For more information about message pricing, see AWS IoT Core Pricing. You can configure the expiration time interval.

Reconnection after a persistent session has expired

If a client doesn't reconnect to its persistent session before it expires, the session ends and its stored messages are discarded. When a client reconnects after the session has expired with a cleanSession flag to 0, the service creates a new persistent session. Any subscriptions or messages from the previous session are not available to this session because they were discarded when the previous session expired.

Persistent session message charges

Messages are charged to your AWS account when the message broker receives a message from a client and sends a message to a client. When a device with a persistent session is not connected, the messages that are stored to be sent when the client reconnects are also charged to your account. When a persistent session expires, messages that have been stored for this session are discarded; however, your account is still charged for their storage. For more information about message pricing, see AWS IoT Core pricing - Messaging.

The default persistent session expiration time of one hour can be increased by using the standard limit increase process. Note that increasing the session expiration time might increase your message charges. The additional time could allow for more messages to be stored for the disconnected device and those additional messages would be charged to your account. The session expiration time is approximate and a session could persist for up to 30 minutes longer than the account limit; however, a session will not be shorter than the account limit. For more information about session limits, see AWS Service Quotas.

Using connectAttributes

ConnectAttributes allow you to specify what attributes you want to use in your connect message in your IAM policies such as PersistentConnect and LastWill. With ConnectAttributes, you can build policies that don't give devices access to new features by default, which can be helpful if a device is compromised.

connectAttributes supports the following features:


Use the PersistentConnect feature to save all subscriptions the client makes during the connection when the connection between the client and broker is interrupted.


Use the LastWill feature to publish a message to the LastWillTopic when a client unexpectedly disconnects.

By default, your policy has non-persistent connection and there are no attributes passed for this connection. You must specify a persistent connection in your IAM policy if you want to have one.

For ConnectAttributes examples, see Connect Policy Examples.

AWS IoT differences from MQTT version 3.1.1 specification

The message broker implementation is based on the MQTT v3.1.1 specification, but it differs from the specification in these ways:

  • AWS IoT supports MQTT quality of service (QoS) levels 0 and 1 only. AWS IoT doesn't support publishing or subscribing with QoS level 2. When QoS level 2 is requested, the message broker doesn't send a PUBACK or SUBACK.

  • In AWS IoT, subscribing to a topic with QoS level 0 means that a message is delivered zero or more times. A message might be delivered more than once. Messages delivered more than once might be sent with a different packet ID. In these cases, the DUP flag is not set.

  • When responding to a connection request, the message broker sends a CONNACK message. This message contains a flag to indicate if the connection is resuming a previous session.

  • When a client subscribes to a topic, there might be a delay between the time the message broker sends a SUBACK and the time the client starts receiving new matching messages.

  • The MQTT specification provides a provision for the publisher to request that the broker retain the last message sent to a topic and send it to all future topic subscribers. AWS IoT doesn't support retained messages. If a request is made to retain messages, the connection is disconnected.

  • The message broker uses the client ID to identify each client. The client ID is passed in from the client to the message broker as part of the MQTT payload. Two clients with the same client ID can't be connected concurrently to the message broker. When a client connects to the message broker using a client ID that another client is using, the new client connection is accepted and the previously connected client is disconnected.

  • On rare occasions, the message broker might resend the same logical PUBLISH message with a different packet ID.

  • The message broker doesn't guarantee the order in which messages and ACK are received.