ActiveMQ supports the Stomp protocol and the Stomp - JMS mapping. This makes it easy to write a client in pure Ruby, Perl, Python or PHP for working with ActiveMQ.

Please see the Stomp site for more details

ActiveMQ v5.6 implements the Stomp v1.1 spec except for allowing spaces at the beginning or end of message header keys, they are preserved in the header values however. In future releases this will not be the case, clients should be updated and user code checked to ensure that spaces in the headers are there intentionally and not as a accident or a client "feature".

Enabling the ActiveMQ Broker for Stomp

To enable STOMP protocol support in the broker add a transport connector definition whose URI scheme is stomp.


   <transportConnector name="stomp" uri="stomp://localhost:61613"/>

To see a full example, try this XML. If you save that XML as foo.xml then you can run stomp via the command line as

activemq xbean:foo.xml

For more help see Run Broker.

The Stomp Wire Format

Stomp uses a text based wire format that can be configured with the following options.  All options can be configured on a Brokers transport bind URI.

Parameter Name

Default Value




Maximum size of the message body (content) that can be sent.



From ActiveMQ 5.12.0: maximum frame size that can be sent. A Stomp frame includes a command, optional headers, and an optional body. Can help help prevent OOM DOS attacks


<transportConnector name="stomp+ssl" uri="stomp+ssl://localhost:61612?wireFormat.maxFrameSize=1000000"/>

Wire format options must have the prefix wireFormat. to take effect, e.g., wireFormat.maxDataLength=100000. Options missing this prefix will be ignored.


From ActiveMQ 5.1: Stomp fully supports ActiveMQ's security mechanism. This means that the CONNECT command will return an ERROR STOMP frame on unsuccessful authentication. Also, the authorization policies will be applied when you try to access (read/write) certain destinations. If you use synchronous operations (by using receipts), you can expect an ERROR frame in case of unauthorized access attempt. In other case, operations will be discarded but the client will not be informed of errors. This applies to all errors that can occur broker-side.

For additional security, you can use Stomp over SSL as described in the following section.

Enabling Stomp over NIO

For better scalability (and performance) you might want to run Stomp protocol over NIO transport. To do that just use stomp+nio transport prefix instead of stomp. For example, add the following transport configuration in your XML file

<transportConnector name="stomp+nio" uri="stomp+nio://localhost:61612"/>

This transport use NIO transport underneath and will generally use much less threads than standard connector. This connector can help if you want to use large number of queues

Stomp NIO connector implementation is available in version 5.3 and up.

Enabling Stomp over SSL

It's easy to configure ActiveMQ to use Stomp over SSL connection. All you have to do is use stomp+ssl transport prefix instead of stomp. For example, add the following transport configuration in your XML file

<transportConnector name="stomp+ssl" uri="stomp+ssl://localhost:61612"/>

Heart-Beat Grace Period

The STOMP protocol (version 1.1 or greater) defines the concept of heart beats as a method by which a client and broker can determine the health of the underlying TCP connection between them.

ActiveMQ offers support for STOMP defined heart beating provided the client is using version 1.1 (or greater) of the protocol. Prior to ActiveMQ 5.9.0, however, the enforcement of the 'read' heart-beat timeout (that is, a heart-beat sent from the client to the broker) was strict. In other words, the broker was intolerant of late arriving read heart-beats from the client. This resulted in the broker concluding that the client was no longer present causing it to close its side of the client's connection when the client failed to honor it's configured heart-beat settings.

As of version 5.9.0 the timeout enforcement for read heart-beats is now configurable via a new transport option transport.hbGracePeriodMultiplier:

   <transportConnector name="stomp" uri="stomp://localhost:61613?transport.hbGracePeriodMultiplier=1.5"/>

This multiplier is used to calculate the effective read heart-beat timeout the broker will enforce for each client's connection. The multiplier is applied to the read-timeout interval the client specifies in its CONNECT frame:

  <client specified read heart-beat interval> * <grace periodmultiplier> == <broker enforced read heart-beat timeout interval>

For backward compatibility, if the grace period multiplier is not configured the default enforcement mode remains strict, e.g., transport.hbGracePeriodMultiplier=1.0. Attempts to configure the grace period multiplier to a value less than, or equal to 1.0 will be silently ignored.

STOMP clients that wish to be tolerant of late arriving heart-beats from the broker must implement their own solution for doing so.

Working with Destinations with Stomp

Note that the prefix in stomp /queue/ or /topic/ is removed from the string before passing it to ActiveMQ as a JMS destination. Also note that the default separator in MOM systems is . (dot). Whilst FOO.BAR is the normal syntax to identify a queue type destination the Stomp equivalent is /queue/FOO.BAR

If in Stomp world you use /queue/foo/bar then in a JMS world the queue would be called foo/bar not /foo/bar.

STOMP messages are non-persistent by default. To use persistent messaging add the following STOMP header to all SEND requests: persistent:true. This default is the opposite of that for JMS messages.

Working with JMS Text/Bytes Messages and Stomp

Stomp is a very simple protocol - that's part of the beauty of it! As such, it does not have knowledge of JMS messages such as TextMessage's or BytesMessage's. The protocol does however support a content-length header. To provide more robust interaction between STOMP and JMS clients, ActiveMQ keys off of the inclusion of this header to determine what message type to create when sending from Stomp to JMS. The logic is simple:

Inclusion of content-length header

Resulting Message





This same logic can be followed when going from JMS to Stomp, as well. A Stomp client could be written to key off of the inclusion of the content-length header to determine what type of message structure to provide to the user.

Message Transformations

The transformation message header on SEND and SUBSCRIBE messages could be used to instruct ActiveMQ to transform messages from text to the format of your desire. Currently, ActiveMQ comes with a transformer that can transform XML/JSON text to Java objects, but you can add your own transformers as well.

Here's a quick example of how to use built-in transformer (taken from test cases)

    private String xmlObject = "<pojo>\n" 
            + "  <name>Dejan</name>\n"
            + "  <city>Belgrade</city>\n" 
            + "</pojo>";

    public void testTransformationReceiveXMLObject() throws Exception {
        MessageProducer producer = session.createProducer(new ActiveMQQueue("USERS." + getQueueName()));
        ObjectMessage message = session.createObjectMessage(new SamplePojo("Dejan", "Belgrade"));
        String frame = "CONNECT\n" + "login: system\n" + "passcode: manager\n\n" + Stomp.NULL;

        frame = stompConnection.receiveFrame();

        frame = "SUBSCRIBE\n" + "destination:/queue/USERS." + getQueueName() + "\n" + "ack:auto" + "\n" + "transformation:jms-object-xml\n\n" + Stomp.NULL;
        frame = stompConnection.receiveFrame();

        frame = "DISCONNECT\n" + "\n\n" + Stomp.NULL;

ActiveMQ uses XStream for its transformation needs. Since it's the optional dependency you have to add it to broker's classpath by putting the appropriate JAR into the lib/ folder. Additionally, if you plan to use JSON transformations you have to add Jettison JSON parser to the classpath.

In order to create your own transformer, you have to do the following:

  1. Build your transformer by implementing a FrameTranslator interface

  2. Associate it with the appropriate header value by creating a file named as a value you want to use in the META-INF/services/org/apache/activemq/transport/frametranslator/ folder of your JAR which will contain the value class=fully qualified classname of your transformer

For example the built-in transformer contains the following value


in the META-INF/services/org/apache/activemq/transport/frametranslator/jms-xml file.


In case you want to debug Stomp communication between broker and clients you should configure the Stomp connector with the trace parameter, like this:

   <transportConnector name="stomp" uri="stomp://localhost:61613?trace=true"/>

This will instruct the broker to trace all packets it sends and receives.

Furthermore, you have to enable tracing for the appropriate log. You can achieve that by adding the following to your conf/

Finally, you will probably want to keep these messages in the separate file instead of polluting the standard broker's log. You can achieve that with the following log4j configuration:

log4j.appender.stomp.layout.ConversionPattern=%d [%-15.15t] %-5p %-30.30c{1} - %m%n, stomp

# Enable these two lines and disable the above two if you want the frame IO ONLY (e.g., no heart beat messages, inactivity monitor etc)., stomp

After this, all your Stomp packets will be logged to the data/stomp.log

Java API

Since version 5.2, there is a simple Java Stomp API distributed with ActiveMQ. Note that this API is provided purely for testing purposes and you should always consider using standard JMS API from Java instead of this one. The following code snippet provides a simple example of using this API:

StompConnection connection = new StompConnection();"localhost", 61613);
connection.connect("system", "manager");
StompFrame connect = connection.receive();

if(!connect.getAction().equals(Stomp.Responses.CONNECTED)) {
	throw new Exception ("Not connected");
connection.send("/queue/test", "message1", "tx1", null);
connection.send("/queue/test", "message2", "tx1", null);
connection.subscribe("/queue/test", Subscribe.AckModeValues.CLIENT);
StompFrame message = connection.receive();
connection.ack(message, "tx2");
message = connection.receive();
connection.ack(message, "tx2");

This example is distributed with the ActiveMQ distribution. You can run it from the example folder with

ant stomp

Stomp Extensions for JMS Message Semantics

Note that STOMP is designed to be as simple as possible - so any scripting language / platform can message any other with minimal effort. STOMP allows pluggable headers on each request such as sending & receiving messages. ActiveMQ has several extensions to the Stomp protocol, so that JMS semantics can be supported by Stomp clients. An OpenWire JMS producer can send messages to a Stomp consumer, and a Stomp producer can send messages to an OpenWire JMS consumer. And Stomp to Stomp configurations, can use the richer JMS message control.

STOMP supports the following standard JMS properties on SENT messages:

STOMP Header

JMS Header




Good consumers will add this header to any responses they send.



Expiration time of the message.



Whether or not the message is persistent.



Priority on the message.



Destination you should send replies to.



Type of the message.



Specifies the Message Groups.



Optional header that specifies the sequence number in the Message Groups.

ActiveMQ Extensions to STOMP

You can add custom headers to STOMP commands to configure the ActiveMQ protocol. Here are some examples:








Should messages be dispatched synchronously or asynchronously from the producer thread for non-durable topics in the broker?

For fast consumers set this to false.

For slow consumers set it to true so that dispatching will not block fast consumers.




I would like to be an Exclusive Consumer on the queue.




For Slow Consumer Handling on non-durable topics by dropping old messages - we can set a maximum-pending limit, such that once a slow consumer backs up to this high water mark we begin to discard old messages.




Specifies whether or not locally sent messages should be ignored for subscriptions.

Set to true to filter out locally sent messages.




Specifies the maximum number of pending messages that will be dispatched to the client. Once this maximum is reached no more messages are dispatched until the client acknowledges a message.

Set to a low value > 1 for fair distribution of messages across consumers when processing messages can be slow

Note: if your STOMP client is implemented using a dynamic scripting language like Ruby, say, then this parameter must be set to 1 as there is no notion of a client-side message size to be sized.

STOMP does not support a value of 0.




Sets the priority of the consumer so that dispatching can be weighted in priority order.




For non-durable topics make this subscription retroactive.




For durable topic subscriptions you must specify the same activemq.client-id on the connection and activemq.subcriptionName on the subscribe prior to v5.7.0.

Note: the spelling subcriptionName NOT subscriptionName. This is not intuitive, but it is how it is implemented in ActiveMQ 4.x.

For the 5.0 release of ActiveMQ, both subcriptionName and subscriptionName will be supported (subcriptionName was removed as of v5.6.0).




Specifies the JMS clientID which is used in combination with the activemq.subcriptionName to denote a durable subscriber.




Specifies a JMS Selector using SQL 92 syntax as specified in the JMS 1.1 specification. This allows a filter to be applied to each message as part of the subscription.