In general, a provider, on its own, is nothing more than the shell for a developer to bundle up some capabilities they wish to dispatch with the command line (or other) clients. It is an analogue to what a "controller" is inside of your MVC application. Before writing your own providers you need to create a manifest, which returns instances of your providers to allow Zend Tool to detect them. A Provider Manifest is an implementation of the Zend_Tool_Framework_Manifest_ProviderManifestable and requires the getProviders() method to return an array of instantiated providers. In anticipation of our first own provider My_Component_HelloProvider we will create the following manifest: A provider manifest class always has to end with the Term "Manifest" otherwise Zend_Tool cannot autodetect it while searching through your include path. As an example, if a developer wants to add the capability of showing the version of a datafile that his 3rd party component is working from, there is only one class the developer would need to implement. Assuming the component is called My_Component, he would create a class named My_Component_HelloProvider in a file named HelloProvider.php somewhere on the include_path. This class would implement Zend_Tool_Framework_Provider_Interface, and the body of this file would only have to look like the following: Given that code above, and assuming the developer wishes to access this functionality through the console client, the call would look like this: As discussed in the architecture section Zend Tool allows to hook different clients for using your Zend Tool providers. To keep compliant with different clients you should use the response object to return messages from your providers instead of using echo or a similiar output mechanism. Rewritting our hello provider with this knowledge it looks like: _registry->getResponse()->appendContent("Hello from my provider!"); } } ]]> As you can see one has to extend the Zend_Tool_Framework_Provider_Abstract to gain access to the Registry which holds the Zend_Tool_Framework_Client_Response instance. The above "Hello World" example is great for simple commands, but what about something more advanced? As your scripting and tooling needs grow, you might find that you need the ability to accept variables. Much like function signatures have parameters, your tooling requests can also accept parameters. Just as each tooling request can be isolated to a method within a class, the parameters of a tooling request can also be isolated in a very well known place. Parameters of the action methods of a provider can include the same parameters you want your client to utilize when calling that provider and action combination. For example, if you wanted to accept a name in the above example, you would probably do this in OO code: The above example can then be called via the command line zf say hello Joe. "Joe" will be supplied to the provider as a parameter of the method call. Also note, as you see that the parameter is optional, that means it is also optional on the command line, so that zf say hello will still work, and default to the name "Ralph". There are cases when the workflow of your provider requires to prompt the user for input. This can be done by requesting the client to ask for more the required input by calling: _registry->getClient()->promptInteractiveInput("Whats your name?"); $name = $name->getContent(); echo 'Hello' . $name . ', from my provider!'; } } ]]> This command throws an exception if the current client is not able to handle interactive requests. In case of the default Console Client however you will be asked to enter the name. Another interesting feature you might wish to implement is pretendability. Pretendabilty is the ability for your provider to "pretend" as if it is doing the requested action and provider combination and give the user as much information about what it would do without actually doing it. This might be an important notion when doing heavy database or filesystem modifications that the user might not otherwise want to do. Pretendability is easy to implement. There are two parts to this feature: 1) marking the provider as having the ability to "pretend", and 2) checking the request to ensure the current request was indeed asked to be "pretended". This feature is demonstrated in the code sample below. _registry->getRequest()->isPretend()) { echo 'I would say hello to ' . $name . '.'; } else { echo 'Hello' . $name . ', from my provider!'; } } } ]]> To run the provider in pretend mode just call: You can also run your provider actions in "verbose" or "debug" modes. The semantics in regard to this actions have to be implemented by you in the context of your provider. You can access debug or verbose modes with: _registry->getRequest()->isVerbose()) { echo "Hello::say has been called\n"; } if($this->_registry->getRequest()->isDebug()) { syslog(LOG_INFO, "Hello::say has been called\n"); } } } ]]> Using the Enviroment variable ZF_CONFIG_FILE or the .zf.ini in your home directory you can inject configuration parameters into any Zend Tool provider. Access to this configuration is available via the registry that is passed to your provider if you extend Zend_Tool_Framework_Provider_Abstract. _registry->getConfig()->username; if(!empty($username)) { echo "Hello $username!"; } else { echo "Hello!"; } } } ]]> The returned configuration is of the type Zend_Tool_Framework_Client_Config but internally the __get and __set magic methods proxy to a Zend_Config of the given configuration type. The storage allows to save arbitrary data for later reference. This can be useful for batch processing tasks or for re-runs of your tasks. You can access the storage in a similar way like the configuration: _registry->getStorage()->get("myUsername"); echo "Hello $aValue!"; } } ]]> The API of the storage is very simple: When designing your providers that are config or storage aware remember to check if the required user-config or storage keys really exist for a user. You won't run into fatal errors when none of these are provided though, since empty ones are created upon request.