atlassian-connect-express: Node.js package for Express.js based Atlassian Add-ons

atlassian-connect-express (ACE) is a toolkit for creating Atlassian Connect based Add-ons with Node.js. Atlassian Connect is a distributed component model for creating add-ons. Add-ons built with Atlassian Connect extend Atlassian cloud-based applications over standard web protocols and APIs. To build add-ons for Atlassian's server (on-premises) products, refer to the server developer documentation.

ACE is the officially supported Node.js framework for Atlassian Connect. Please read our documentation to see the other supported and community provided Frameworks and Tools. You will find the recommended tools extremely useful when writing your own Atlassian Connect add-ons; be sure to peruse the list of tools and use them as much as possible to aid development.

Please ensure you always use the latest patch version of ACE to ensure your add-on has the latest security patches and fixes. Versions prior to 1.0.14 and 2.0.2 have a known security vulnerability.

More info

ACE helps you get started developing add-ons quickly, using Node.js and Express as the add-on server.

It's important to understand that Express by itself is a web app framework for Node. ACE just provides a library of middleware and convenience helpers that make it easier to build Atlassian add-ons. Specifically, ACE adds:

• An optimized dev loop by handling registration and deregistration on the target Atlassian application for you at startup and shutdown.
• A filesystem watcher that detects changes to atlassian-connect.json. When changes are detected, the add-on is re-registered with the host(s).
• Automatic JWT authentication of inbound requests as well as JWT signing for outbound requests back to the host.
• Automatic persistence of host details (i.e., client key, host public key, host base url etc.).
• ngrok tunnel, for testing with Jira or Confluence hosts.

Release notes

For detailed release notes and upgrade guides, please see the Release Notes.

Getting started

The fastest way to get started is to install the atlas-connect CLI tool. The CLI makes it possible to generate an ACE enabled add-on scaffold very quickly. To install:

npm i -g atlas-connect

Create a project

Let's start by creating an add-on project:

atlas-connect new <project_name>

This creates a new project in the current directory.

Install dependencies

Change to the new project directory and install dependencies:

npm install

If you get any errors related to node-gyp (especially with Node 8 on Windows), try installing its prerequisites.

Setting up a development environment

At this point, you're all set to run your add-on, but you still need to install it in Jira or Confluence. You can install your new add-on in any Jira or Confluence host where you are an administrator, but usually it's best to create a new host for you to use during development. Follow this link to sign up for a free Jira or Confluence Cloud host.

Running your add-on server

In your project directory, run:

npm start

This will boot up your Express server on the default port of 3000.

Dev loop

At this point, you can start building your add-on. Changes to views load automatically, however, if you make changes to any JavaScript, you need to restart Express. If you want your server to automatically restart when your JavaScript changes, consider using nodemon or the like.

Automatic registration

This section will describe how to configure ACE so that it can automatically register your add-on with your Jira or Confluence host, re-register on changes to the descriptor, and de-register on shut down.

To get this functionality, you will need to:

Generate an Atlassian API token

• Create a file called credentials.json,
• Copy and paste the contents of this file,
• Add credentials.json to the .gitignore file, and
• Change the contents of the file to contain the link to your development environment, e-mail as the username, API token as password and product name.

Authorise and Configure ngrok (Version 10.0.0+)

• Sign up for an account: https://dashboard.ngrok.com/signup
• Obtain an auth token: https://dashboard.ngrok.com/get-started/your-authtoken
• Add it to a top-level field in credentials.json like this "ngrok": {"authtoken": "your_token"}. It accepts all valid ngrok parameters.

ACE will now read this file and automatically create an ngrok tunnel, and register your add-on on your development host.

Configuration

The configuration for your add-on is done in two files:

• ./config.json -- This file contains the configuration for each runtime environment your plugin runs in. The file has comments to help you understand available settings.
• ./atlassian-connect.json -- This file is a manifest of all the extension points your add-on uses. To see all of the available extension point options for Jira, check out the modules sections of the atlassian-connect Jira modules documentation. To check out all the available extension point options for Confluence, check out the modules section of the atlassian-connect Confluence modules documentation.

The behaviour of your add-on can be further configured by setting the AC_OPTS environment variable (see the end of this section).

config.json

The ./config.json file contains all of the settings for the add-on server. This file is broken into environments.

{
  // This is the default environment. To change your app to use
  // a different env, set NODE_ENV (http://expressjs.com/api.html#app.configure)
  "development": {
    // Set to true if your app contains a errorHandler middleware:
    // https://expressjs.com/en/guide/error-handling.html
    "expressErrorHandling": false,

    // This is the port your Express server will listen on
    "port": 3000,

    // Use views/unauthorized.hbs for error page.
    "errorTemplate": true,
    // Use custom error template
    // "errorTemplateName": "auth-response-handler.jsx",
    
    // Use Custom error object for error template
    // "errorTemplateObject": {
    //     "title": "App not authorized!",
    //     "unauthorized": true
    // },

    // To enable validation of descriptor on startup and every time it changes,
    // add the optional config validateDescriptor to true
    "validateDescriptor": true,

    // atlassian-connect-express currently integrates with Sequelize for
    // persistence to store the host client information (i.e., client key,
    // host public key, etc). When no adapter is specified, it defaults to
    // Sequelize's fallback memory storage.
    //
    // To specify a backend for Sequelize  other than "memory", set the
    // "dialect" value to one of Sequelize's other supported dialects.
    //
    // To use your own storage adapter, add the key
    // "adapter" to the following configuration, and replace "dialect"
    // and "connection" with any values your adapter expects.  Then make sure
    // that you register your adapter factory with the following code in
    // app.js:
    //
    //   ac.store.register(adapterName, factoryFn)
    //
    // See `atlassian-connect-express/lib/store/index.js` and the default
    // `sequelize.js` files for code demonstrating how to write a
    // conformant adapter.  The default values are as follows:
    //
    //   "store": {
    //     "adapter": "sequelize",
    //     "dialect": "sqlite",
    //     "storage": ":memory:"
    //   },
    //
    //
    // To use sqlite with a persistent file instead of an in-memory database, the following settings can be used,
    // and will result in a store.db file being created in the working directory of the app:
    // 
    //   "store": {
    //     "adapter": "sequelize",
    //     "dialect": "sqlite3",
    //     "url": "sqlite://./store.db"
    //   },
    //
    // To instead configure, say, a PostgreSQL store, the following could be
    // used:
    //
    //   "store": {
    //     "adapter": "sequelize",
    //     "dialect": "postgres",
    //     "url": "postgres://localhost/my_addon_database",
    //     "table": "MySetting" //optional - table name. Default is "AddonSetting"
    //     "logging": function, //optional - function that gets executed every time Sequelize would log something.
    //     "pool": {}           //optional - pool options you have that you may pass to sequelize adapter
    //   },
    //
    // For MongoDB, use the following:
    //
    //   "store": {
    //     "adapter": "mongodb",
    //     "url": "mongodb://localhost:27017/my_addon_database",
    //     "collection": "AddonSettings"
    //   },
    //
    // For Redis, use the following:
    //
    //   "store": {
    //     "adapter": "redis",
    //     "url": "redis://localhost:6379",
    //   },    
    //
    // For DynamoDB, use the following:
    //
    //   "store": {
    //     "adapter": "dynamodb",
    //     "table": "AddonSettings",
    //     "region": "us-east-1",
    //   },
    //
    //  Note that the DynamoDB table should have "clientKey" as a string
    //  partition key and "key" as a string sort key.
    //
    // You will also need an appropriate Sequelize driver if you choose something
    // other than the default "dialect".  In the PostgreSQL case you'd need to
    // run the following command to add the proper support:
    //
    //   $ npm install --save pg
    //
    // You will also need aws-sdk if you choose the DynamoDB driver:
    //
    //   $ npm install --save @aws-sdk/client-dynamodb
    //

  },

  // This is the production add-on configuration, which is enabled by setting
  // the NODE_ENV=production environment variable.
  "production": {
    // On a PaaS host like Heroku, the runtime environment will provide the
    // HTTP port to you via the PORT environement variable, so we configure
    // that to be honored here.
    "port": "$PORT",
    // This is the public URL to your production add-on.
    "localBaseUrl": "https://your-subdomain.herokuapp.com",
    // This is the URL to atlassian connect all.js.
    "hostScriptUrl": "https://connect-cdn.atl-paas.net/all.js",
    "store": {
      // You won't want to use the memory store in production, or your install
      // registrations will be forgotten any time your app restarts.  Here
      // we tell atlassian-connect-express to use the PostgreSQL backend for the default
      // Sequelize adapter.
      "dialect": "postgres",
      // Again, a PaaS host like Heroku will probably provide the db connection
      // URL to you through the environment, so we tell atlassian-connect-express to use that value.
      "url": "$DATABASE_URL"
    },

    // If your app supports multiple baseUrls,
    // additional baseUrls can be added here so that it can be used to
    // verify the `audience` claim during installation lifecycle callback.
    "allowedBaseUrls" : [
      "https://other-domain.herokuapp.com"
    ],

    // Make sure that your add-on can only be registered by the hosts on
    // these domains.
    "whitelist": [
      "*.jira-dev.com",
      "*.atlassian.net",
      "*.jira.com"
    ]
  }
}

AC_OPTS

The AC_OPTS environment variable can be used to change the behaviour of ACE for ease of development, like so:

AC_OPTS=no-auth,force-reg node app.js

Set it to a space or comma delimited list containing one or more of the following values.

force-reg Make the add-on always register itself with running a Jira or Confluence host when it starts up (normally auto-registration only happens if the add-on is using a memory store).

force-dereg Make the add-on always de-register itself with running a Jira or Confluence host on shutdown (normally auto-registration only happens if the add-on is using a memory store or running in development mode).

no-reg Make the add-on never register itself with running a Jira or Confluence host (i.e. don't auto-register even if a memory store is being used).

no-auth Skip authentication of incoming requests (i.e. don't check for or validate JWT tokens).

atlassian-connect.json

The atlassian-connect.json describes what your add-on will do. There are three main parts to the descriptor: meta information that describes your add-on (i.e., name, description, key, etc.), permissions and authentication information, and a list of the components your add-on will extend. This descriptor is sent to the host (i.e., Jira or Confluence) when your add-on is installed.

To see all of the available settings in the atlassian-connect.json, visit the module sections of the atlassian-connect documentation

If you need a pre-processing step to your descriptor, you can configure one by changing your app.js so that a transformer is included in the config. The descriptorTransformer property expects to be a function and passes in descriptor as an object, and the app.config object.

var addon = ac(app, { config: {
  descriptorTransformer: function(descriptor, config) {
    if (config.environment() === "production") {
      descriptor.key = "production-key";
    }
    return descriptor;
  }
}});

Sample Add-ons

• Jira Example -- a simple Jira example add-on
• Confluence Example -- a simple Confluence example add-on

Scaffold

When you generate a new ACE add-on, you're actually just downloading a copy of the Atlassian Connect for Express.js template.

Handlebars layouts and templates

The base scaffold uses the Handlebars template library via the express-hbs package.

Handlebars views are stored in the ./views directory. The base template contains a layout.hbs and a sample page (hello-world.hbs). Handlebars alone doesn't provide layouts, but the express-hbs package does. To apply the layout.hbs layout to your template page, just add the following to the top of your template:

{{!< layout}}

To learn more about how Handlebars works in express.js, take a look at the express-hbs documentation.

Special context variables

ACE injects a handful of useful context variables into your render context. You can access any of these within your templates:

• title: the add-on's name (derived from atlassian-connect.json)
• addonKey: the add-on key defined in atlassian-connect.json
• localBaseUrl: the base URI of the add-on
• hostBaseUrl: the base URI of the target application (includes the context path if available)
• hostStylesheetUrl: the URL to the base CSS file for Connect add-ons. This stylesheet is a bare minimum set of styles to help you get started. It's not a full AUI stylesheet.
• hostScriptUrl: the URL to the Connect JS client. This JS file contains the code that will establish the seamless iframe bridge between the add-on and its parent. It also contains a handful of methods and objects for accessing data through the parent (look for the AP JS object).
• token: the token that can be used to authenticate calls from the iframe back to the add-on service.
• license: the license status
• context: the JWT context claim
• clientKey: the client consumer key used to identity the Jira or Confluence host from which the request came
• userAccountId: the Atlassian Account ID of the user.
• userId: (deprecated) the username of the user from which the request came.
• timeZone: (deprecated) the user's timezone
• locale: (deprecated) the user's locale

You can access any of the variables above as normal Handlebars variables. For example, to generate a link in your page that links elsewhere in the host:

<a href="{{hostBaseUrl}}/browse/JRA">Jira</a>

Events emitted

All products

• host_settings_saved: after /installed lifecycle, ACE tries to save the client information (baseUrl, clientKey, app key, puglinsVersion, productType, publicKey, serverVersion, sharedSecret) in storage. If successfuly saved, this event is emitted
• host_settings_not_saved: after /installed lifecycle, ACE tries to save the client information in storage. If there's any error or problem, this event is emitted
• addon_registered: after an ngrok tunnel is created, ACE will try to register or install the app in a Jira or Confluence product
• webhook_auth_verification_triggered: ACE automatically registers webhooks and corresponding paths in the descriptor file, once it tries to authenticate this event is emitted
• webhook_auth_verification_successful: ACE automatically registers webhooks and corresponding paths in the descriptor file, once it tries to authenticate and is successful, this event is emitted

Jira/Confluence

• localtunnel_started: event emitted after ACE successfully creates an ngrok tunnel
• addon_deregistered: when ACE receives a SIGTERM, SIGINT, and SIGUSR2 signals, it will deregister the app and this event is emitted

To listen to an event:

addon.on(event, function() {
  // Add something here
});

Recipes

How to secure a route with JWT

Add-ons are authenticated through JWT. To simplify JWT verification on your routes, you can simply add an ACE middleware to your route:

module.exports = function(app, addon) {
  app.get(
    '/protected-resource',

    // Protect this resource with JWT
    addon.authenticate(),

    function(req, res) {
      res.render('protected');
    }
  );
};

Simply adding the addon.authenticate() middleware will protect your resource.

How to secure an installation callback route with JWT

Install/Uninstall hooks are authenticated through asymmetric JWT. You can simply add an ACE authenticateInstall middleware to your custom install or uninstall lifecycle route:

module.exports = function(app, addon) {
  app.post(
    '/uninstalled',

    // Protect this install callback resource with asymmetric JWT
    addon.authenticateInstall(),

    function(req, res) {
      res.status(204).send();
    }
  );
};

Authorizing requests

If your app API accepts requests from the front-end using context JWTs, it's important to perform authorization checks. Atlassian Connect Express includes a basic middleware that leverages the Jira get bulk permissions API and Confluence check content permissions API to perform authorization checks based on the context JWT.

For example to restrict an endpoint to Jira admins that are also project admins:

module.exports = function(app, addon) {
  app.get(
    '/admin-resource',

    [
      addon.authenticate(true /* accept context JWTs */),
      // only allow product admins that are also project admins
      addon.authorizeJira({ global: ["ADMINISTER"], project: ["ADMINISTER_PROJECTS"] })
    ],

    function(req, res) {
      res.render('protected');
    }
  );
};

Similarly for Confluence, to restrict an endpoint to Confluence admins that also have permissions to read the current page:

module.exports = function(app, addon) {
  app.get(
    '/admin-resource',

    [
      addon.authenticate(true /* accept context JWTs */),
      // only allow product admins that also have access to read the current page
      addon.authorizeConfluence({ application: ["administer"], content: "read" })
    ],

    function(req, res) {
      res.render('protected');
    }
  );
};

How to send a signed HTTP request from the iframe back to the add-on service

The initial call to load the iframe content is secured by JWT, as described above. However, the loaded content cannot sign subsequent requests. A typical example is content that makes AJAX calls back to the add-on. Cookie sessions cannot be used, as many browsers block third-party cookies by default. ACE provides middleware that works without cookies and helps making secure requests from the iframe.

Standard JWT tokens are used to authenticate requests from the iframe back to the add-on service. A route can be secured using the addon.checkValidToken() middleware:

module.exports = function(app, addon) {
  app.get(
    '/protected-resource',

    // Require a valid token to access this resource
    addon.checkValidToken(),

    function(req, res) {
      res.render('protected');
    }
  );
};

In order to secure your route, the token must be part of the HTTP request back to the add-on service. This can be done by using the standard jwt query parameter:

<a href="/protected-resource?jwt={{token}}">See more</a>

The second option is to use the Authorization HTTP header, e.g. for AJAX requests:

beforeSend: function(request) {
  request.setRequestHeader("Authorization", "JWT {{token}}");
}

You can embed the token anywhere in your iframe content using the token content variable. For example, you can embed it in a meta tag, from where it can later be read by a script:

<meta name="token" content="{{token}}">

How to send a signed outbound HTTP request back to the host

ACE bundles and extends the request HTTP client. To make a JWT signed request back to the host, all you have to do is use request the way it was designed, but use a URL back to the host's REST APIs.

var httpClient = addon.httpClient(req);
httpClient.get('/', function(err, res, body) {
  ...
});

If not in a request context, you can perform the equivalent operation as follows:

var httpClient = addon.httpClient({
  clientKey: clientKey  // The unique client key of the tenant to make a request to
});
httpClient.get('/', function(err, res, body) {
  ...
});

By default, these requests are authenticated as the add-on. If you would like to make a request as a specific user, the #asUserByAccountId() method should be used. Under the covers, an OAuth2 bearer token will be retrieved for the user you've requested.

var httpClient = addon.httpClient(req);
httpClient.asUserByAccountId('ebcab857-c769-4fbd-8ad6-469510a43b87').get('/rest/api/latest/myself', function(err, res, body) {
  ...
});

Ensure you pass the userAccountId value into the method, and not the username or userKey. If you were previously using #asUser() with userKey, you can convert it into a userAccountId through the User REST resource on the host product.

• Jira
• Confluence

You can also set custom headers or send a form data. Take, for example this request which attaches a file to a Jira issue:

var filePath = path.join(__dirname, 'some.png');
fs.readFile(filePath, function(err, data) {
  httpClient.post({
    url: '/rest/api/2/issue/' + issueKey + '/attachments',
    headers: {
      'X-Atlassian-Token': 'nocheck'
    },
    multipartFormData: {
      file: [data, { filename: 'some.png' }]
    }
  },
  function(err, httpResponse, body) {
    if (err) {
      return console.error('Upload failed:', err);
    }
    console.log('Upload successful:', body);
  });
});

Using the product REST API

Certain REST URLs may require additional scopes that should be added to your atlassian-connect.json file.

Forge

Your ACE app can act as a remote backend for Forge.

To do so, it must have an appId set in the ./config.json. The value should be the app.id as it appears in the Forge manifest. For example:

{
  "appId": "ari:cloud:ecosystem::app/406d303d-0393-4ec4-ad7c-1435be94583a"
}

If you are using the DynamoDB adapter for storage, you will also need to add a couple of tables for Forge installation related data. There is code that will attempt to create these tables here but this may not succeed if your app does not have the necessary permissions. In that case please manually create the following:

• An InstallationClientKeys table with installationId as a string partition key.

• A ForgeSettings table with installationId as a string partition key, and key as a string sort key.

Authenticating requests from Forge

The addon.authenticateForge() middleware will verify remote requests from Forge in accordance with the Remote Contract and set the following properties on request.context.forge. Refer to the remote contract reference for more information about these properties.

• request.context.forge.tokens.app: A ready-to-use access token representing the app
• request.context.forge.tokens.user: A ready-to-use access token representing the current user
• request.context.forge.apiBaseUrl: The API base URL where all product API requests should be routed. Example: https://api.atlassian.com/ex/confluence/4c822e2f-510f-48b9-b8d2-8419d0932949
• request.context.forge.app: Information about the app and installation context
• request.context.forge.context: Product- and extension-specific data including configuration.

Making requests

There is no pre-configured http client for Forge akin to the Connect addon.httpClient, but making a Forge-authenticated request in your http client of choice is as simple as prepending the apiBaseUrl and providing one of the ready-to-use tokens in the Authorization header, prefixed by Bearer . For example:

app.get("/forge/backend/resource", addon.authenticateForge(), async (req, res) => {
  const response = await fetch(`${req.context.forge.apiBaseUrl}/wiki/api/content`, {
    headers: {
      Authorization: `Bearer ${req.context.forge.tokens.app}`
    }
  });
});

The addon.authenticateForge() middleware caches the app token and API base URL against the installation ID for later use. If you are making a request in response to an invocation which did not come from Forge, have the appropriate installation ID, and would like to make a request to Atlassian APIs, you can retieve the app token using addon.getForgeAppToken(installationId). For example:

  app.get("/external-event/{installationId}", async (req, res) => {
    const { appToken, apiBaseUrl } = await addon.getForgeAppToken(req.params.installationId);
    const response = await fetch(`${apiBaseUrl}/wiki/api/content`, {
      headers: {
        Authorization: `Bearer ${appToken}`
      }
    });
  });

Associating Connect and Forge data

To have access to Connect context variables and http client as well as Forge context variables, the addon.associateConnect() middleware can be used in tandem with addon.authenticateForge():

app.get("/forge/backend/resource", [addon.authenticateForge(), addon.associateConnect()], async (req, res) => {
  /* Both Connect and Forge context are available here, for example:
  addon.httpClient(req);
  addon.settings.set("key", { value: "val" }, req.context.clientKey);
  req.context.forge;
  */
});

How to deploy to Heroku

Before you start, install Git and the Heroku Toolbelt.

If you aren't using git to track your add-on, now is a good time to do so as it is required for Heroku. Ensure you are in your project home directory and run the following commands:

git config --global user.name "John Doe"
git config --global user.email [email protected]
ssh-keygen -t rsa
git init
git add .
git commit . -m "some message"
heroku keys:add

Next, create the app on Heroku:

heroku apps:create <add-on-name>

Next, let's store our registration information in a Postgres database. In development, you were likely using the memory store. In production, you'll want to use a real database.

heroku addons:add heroku-postgresql:hobby-dev --app <add-on-name>

Lastly, let's add the project files to Heroku and deploy!

If you aren't already there, switch to your project home directory. From there, run these commands:

git remote add heroku [email protected]:<add-on-name>.git
git push heroku master

It will take a minute or two for Heroku to spin up your add-on. When it's done, you'll be given the URL where your add-on is deployed, however, you'll still need to register it on your Jira or Confluence host.

In order to run your add-on on a Jira or Confluence host, you must enter production mode. To achieve this, set the NODE_ENV variable to production like so:

heroku config:set NODE_ENV=production

For further detail, we recommend reading Getting Started with Node.js on Heroku.

Troubleshooting

"Unable to connect and retrieve descriptor from http://localhost:3000/atlassian-connect.json, message is: java.net.ConnectException: Connection refused"

You'll get this error if Jira or Confluence can't access http://localhost:3000/atlassian-connect.json. One way to debug this is to see what the command hostname returns.

If it returns localhost, change it. On a OS X, you'll need to set a proper "Computer Name" in System Preferences > Sharing.

Debugging HTTP Traffic

Several tools exist to help snoop the HTTP traffic between your add-on and the host server:

• Enable node-request's HTTP logging by starting your app with NODE_DEBUG=request node app
• Check out the HTTP-debugging proxies Charles and Fiddler
• Try local TCP sniffing with justniffer by running something like justniffer -i eth0 -r, substituting the correct interface value

Getting help

If you need help using Express.js, see the API reference or developer's guide.

If you need help developing against Atlassian products, see the Atlassian Developer site.

If you need help using functionality provided by ACE, or would like to report a problem with the toolkit, please post in the Atlassian Developer Community.

Contributing

Pull requests, issues, and comments are welcome! It's also open source Apache 2.0. So, please feel free to fork and send us pull requests.

For pull requests:

• Add tests for new features and bug fixes.
• Follow the existing style.
• Separate unrelated changes into multiple pull requests. See the existing issues for things to start contributing.

For bigger changes, make sure you start a discussion first by creating an issue and explaining the intended change.

Pull Requests

We take the 'trust' approach when it comes to pull requests. This means that reviewers should feel comfortable approving a PR with outstanding items due to trust in knowing that the PR owner will address the actions as necessary. Use actions for PR items that are particularly important since these require the PR owner to explicitly acknowledge them before merging

Unit tests

Run npm run test.

Note: the MongoDB tests will fail locally for M1 macbooks due to a missing binary which is why we force mongodbMemoryServer to use x64 architecture.