Elixir Plugs

A cross-post collaboration with Elixir School

This blog post was originally published on Elixir School as a lesson on Plug. I added the section on Plug.ErrorHandler and Sophie DeBenedetto brought the lesson up to date to use Cowboy2 protocol. This lesson is, of course, a collaboration with many other open source contributors over at Elixir School. Please check out the source of the lesson for the full list of contributors, and feel free to put in a pull request to help us keep improving Elixir School for others! And without further interruption, on to the lesson…

If you’re familiar with Ruby, you can think of Plug as Rack with a splash of Sinatra. It provides a specification for web application components and adapters for web servers. While not part of Elixir core, Plug is an official Elixir project.

In this lesson, we’ll build a simple HTTP server from scratch using the PlugCowboy Elixir library. Cowboy is a simple HTTP server for Erlang and Plug will provide us with a connection adapter for that web server.

After we set up our minimal web application, we’ll learn about Plug’s router and how to use multiple plugs in a single web app.

Table of Contents


This tutorial assumes you have Elixir 1.5 or higher and mix installed already.

We’ll start by creating a new OTP project, with a supervision tree.

We need our Elixir app to include a supervision tree because we will use a Supervisor to start up and run our Cowboy2 server.


Adding dependencies is a breeze with mix. To use Plug as an adapter interface for the Cowboy2 webserver, we need to install the PlugCowboy package:

Add the following to your mix.exs file:

At the command line, run the following mix task to pull in these new dependencies:

The Plug Specification

In order to begin creating Plugs, we need to know, and adhere to, the Plug spec. Thankfully for us, there are only two functions necessary: init/1 and call/2.

Here’s a simple Plug that returns “Hello World!”:

Save the file to lib/example/hello_world_plug.ex.

The init/1 function is used to initialize our Plug’s options. It is called by a supervision tree, which is explained in the next section. For now, it’ll be an empty List that is ignored.

The value returned from init/1 will eventually be passed to call/2 as its second argument.

The call/2 function is called for every new request that comes in from the web server, Cowboy. It receives a %Plug.Conn{} connection struct as its first argument and is expected to return a %Plug.Conn{} connection struct.

Configuring the Project’s Application Module

We need to tell our application to start up and supervise the Cowboy web server when the app starts up.

We’ll do so with the Plug.Cowboy.child_spec/1 function.

This function expects three options:

  • :scheme — HTTP or HTTPS as an atom (:http:https)
  • :plug — The plug module to be used as the interface for the web server. You can specify a module name, like MyPlug, or a tuple of the module name and options {MyPlug, plug_opts}, where plug_opts gets passed to your plug modules init/1 function.
  • :options — The server options. Should include the port number on which you want your server listening for requests.

Our lib/example/application.ex file should implement the child spec in its start/2 function:

Note: We do not have to call child_spec here, this function will be called by the supervisor starting this process. We simply pass a tuple with the module that we want the child spec built for and then the three options needed.

This starts up a Cowboy2 server under our app’s supervision tree. It starts Cowboy running under the HTTP scheme (you can also specify HTTPS), on the given port, 8080, specifying the plug, Example.HelloWorldPlug, as the interface for any incoming web requests.

Now we’re ready to run our app and send it some web requests! Notice that, because we generated an OTP app with the —- sup flag, our Example application will start up automatically thanks to the application function.

In mix.exs, you should see the following:

We’re ready to try out this minimalistic, plug-based web server. On the command line, run: $ mix run --no-halt

Once everything is finished compiling, and [info] Starting application… appears, open a web browser to It should display:

Hello World!


For most applications, like a web site or REST API, you’ll want a router to route request for different paths and HTTP verbs to different handlers.
Plug provides a router to do that. As we are about to see, we don’t need a framework like Sinatra in Elixir since we get that for free with Plug.

To start let’s create a file at lib/example/router.ex and copy the following into it:

This is a bare minimum Router but the code should be pretty self-explanatory.
We’ve included some macros through use Plug.Router and then set up two of the built-in Plugs: :match and :dispatch.
There are two defined routes, one for handling GET requests to the root and the second for matching all other requests so we can return a 404 message.

Back in lib/example.ex, we need to add Example.Router into the web server supervisor tree. Swap out the Example.HelloWorldPlug plug with the new router:

Start the server again, stopping the previous one if it’s running (press Ctrl+C twice).

Now in a web browser, go to It should output Welcome. Then, go to, or any other path. It should output Oops! with a 404 response.

Adding Another Plug

It is common to use more than one plug in a given web application, each of which is dedicated to its own responsibility. For example, we might have a plug that handles routing, a plug that validates incoming web requests, a plug that authenticates incoming requests, etc. In this section, we’ll define a plug to verify incoming requests parameters and we’ll teach our application to use both of our plugs — the router and the validation plug.

We want to create a Plug that verifies whether or not the request has some set of required parameters. By implementing our validation in a Plug we can be assured that only valid requests will make it through to our application. We will expect our Plug to be initialized with two options: :paths and :fields. These will represent the paths we apply our logic to and which fields to require.

Note: Plugs are applied to all requests which is why we will handle filtering requests and applying our logic to only a subset of them.
To ignore a request we simply pass the connection through.

We’ll start by looking at our finished Plug and then discuss how it works. We’ll create it at lib/example/plug/verify_request.ex:

The first thing to note is we have defined a new exception IncompleteRequestError and that one of its options is :plug_status. When available this option is used by Plug to set the HTTP status code in the event of an exception.

The second portion of our Plug is the call/2 function. This is where we decide whether or not to apply our verification logic. Only when the request’s path is contained in our :paths option will we call verify_request!/2.

The last portion of our plug is the private function verify_request!/2 which verifies whether the required :fields are all present. In the event that some are missing, we raise IncompleteRequestError.

We’ve set up our Plug to verify that all requests to /upload include both “content” and “mimetype”. Only then will the route code be executed.

Next, we need to tell the router about the new Plug. Edit lib/example/router.ex and make the following changes:

With this code, we are telling our application to send incoming requests through the VerifyRequest plug before running through the code in the router. Via the function call:

We automatically invoke VerifyRequest.init(fields: [“content”, “mimetype”], paths: [“/upload”]). This in turn passes the given options to the VerifyRequest.call(conn, opts) function.

Let’s take a look at this plug in action! Go ahead and crash your local server (remember, that’s done by pressing ctrl + c twice). Then restart the server (mix run --no-halt). Now go to in your browser and you’ll see that the page simply isn’t working. We’re not even getting our ‘Oops!’ message. Now let’s add our required params by going to Now we should see our ‘Uploaded’ message. It’s not great that when we throw an error we don’t get any page, but we’ll deal with how to handle errors with plugs later.

Making The HTTP Port Configurable

Back when we defined the Example module and application, the HTTP port was hard-coded in the module. It’s considered good practice to make the port configurable by putting it in a configuration file.

We’ll set an application environment variable in config/config.exs

Next we need to update lib/example/application.ex read the port configuration value, and pass it to Cowboy. We’ll define a private function to wrap up that responsibility

The third argument of Application.get_env is the default value, for when the configuration directive is undefined.

Now to run our application we can use: $ mix run --no-halt.

Testing a Plug

Testing Plugs is pretty straightforward thanks to Plug.Test. It includes a number of convenience functions to make testing easy.

Write the following test to test/example/router_test.exs:

Run it with this: $ mix test test/example/router_test.exs


We noticed earlier that when we go to we don’t even see an error page. Let’s fix that now by adding in Plug.ErrorHandler.

First, open up lib/example/router.ex and then write the following to that file.

You’ll notice at the top we are now adding use Plug.ErrorHandler. This plug now catches any error and then looks for a function handle_errors/2 to call. handle_errors just needs to accept the conn as the first argument and then a map with three items (:kind:reason, and :stack) as the second.
You can see we’ve defined a very some handle_error to see what’s going on. Let’s stop and restart our app again to see how this works!

Now, when you navigate to, we see an error message ‘Something went wrong’. If you look in your terminal, you’ll see something like the following:

This plug makes it really easy to catch the useful information needed for developers to fix issues while being able to also give our end user a nice page so it doesn’t look like our app totally blew up!

Available Plugs

There are a number of Plugs available out-of-the-box. The complete list can be found in the Plug docs here.

Thanks for reading! Want to work on a mission-driven team that loves Elixir and contributing to open source projects? We’re hiring!

Footer top

To learn more about Flatiron School, visit the website, follow us on Facebook and Twitter, and visit us at upcoming events near you.

Flatiron School is a proud member of the WeWork family. Check out our sister technology blogs WeWork Technology and Making Meetup.

Footer bottom