fastify-grant

Fastify plugin for Grant OAuth Proxy

var fastify = require('fastify')
var cookie = require('fastify-cookie')
var session = require('fastify-session')
var grant = require('fastify-grant')

fastify()
  .register(cookie)
  .register(session, {secret: 'grant', cookie: {secure: false}})
  .register(grant({/*configuration - see below*/}))
  .listen(3000)

Examples

ES Modules and TypeScript

Import Grant in your .mjs files:

import fastify from 'fastify'
import cookie from 'fastify-cookie'
import session from 'fastify-session'
import grant from 'fastify-grant'

fastify()
  .register(cookie)
  .register(session, {secret: 'grant', cookie: {secure: false}})
  .register(grant({/*configuration - see below*/}))
  .listen(3000)

Grant ships with extensive type definitions for TypeScript. However, a few additional type definitions that extend the typings of your HTTP framework of choice can be found here.

Configuration

Configuration: Basics

{
  "defaults": {
    "origin": "http://localhost:3000",
    "transport": "session",
    "state": true
  },
  "google": {
    "key": "...",
    "secret": "...",
    "scope": ["openid"],
    "nonce": true,
    "custom_params": {"access_type": "offline"},
    "callback": "/hello"
  },
  "twitter": {
    "key": "...",
    "secret": "...",
    "callback": "/hi"
  }
}

• defaults - default configuration for all providers
  • origin - where your client server can be reached http://localhost:3000 | https://site.com ...
  • transport - a transport used to deliver the response data in your callback route
  • state - generate random state string
• provider - any supported provider google | twitter ...
  • key - consumer_key or client_id of your OAuth app
  • secret - consumer_secret or client_secret of your OAuth app
  • scope - array of OAuth scopes to request
  • nonce - generate random nonce string (OpenID Connect only)
  • custom_params - custom authorization parameters
  • callback - relative route or absolute URL to receive the response data /hello | https://site.com/hey ...

Configuration: Description

Configuration: Values

Connect

Connect: Origin

{
  "defaults": {
    "origin": "http://localhost:3000"
  }
}

The origin is where your client server can be reached.

You login by navigating to the /connect/:provider route where :provider is a key in your configuration, usually one of the officially supported ones, but you can define your own as well. Additionally you can login through a static override defined for that provider by navigating to the /connect/:provider/:override? route.

Connect: Prefix

By default Grant operates on the following two routes:

/connect/:provider/:override?
/connect/:provider/callback

However, the default /connect prefix can be configured:

{
  "defaults": {
    "origin": "http://localhost:3000",
    "prefix": "/oauth"
  }
}

Connect: Redirect URI

The redirect_uri of your OAuth app should follow this format:

[origin][prefix]/[provider]/callback

Where origin and prefix have to match the ones set in your configuration, and provider is a provider key found in your configuration.

For example: http://localhost:3000/connect/google/callback

This redirect URI is used internally by Grant. Depending on the transport being used you will receive the response data in the callback route or absolute URL configured for that provider.

Connect: Custom Parameters

Some providers may employ custom authorization parameters that you can configure using the custom_params key:

{
  "google": {
    "custom_params": {"access_type": "offline", "prompt": "consent"}
  },
  "reddit": {
    "custom_params": {"duration": "permanent"}
  },
  "trello": {
    "custom_params": {"name": "my app", "expiration": "never"}
  }
}

Connect: OpenID Connect

The openid scope is required, and generating a random nonce string is optional but recommended:

{
  "google": {
    "scope": ["openid"],
    "nonce": true
  }
}

Grant does not verify the signature of the returned id_token by default.

However, the following two claims of the id_token are being validated:

1. aud - is the token intended for my OAuth app?
2. nonce - does it tie to a request of my own?

Connect: PKCE

PKCE can be enabled for all providers or for a specific provider only:

{
  "google": {
    "pkce": true
  }
}

Providers that do not support PKCE will ignore the additional parameters being sent.

Connect: Static Overrides

Provider sub configurations can be configured using the overrides key:

{
  "github": {
    "key": "...", "secret": "...",
    "scope": ["public_repo"],
    "callback": "/hello",
    "overrides": {
      "notifications": {
        "key": "...", "secret": "...",
        "scope": ["notifications"]
      },
      "all": {
        "scope": ["repo", "gist", "user"],
        "callback": "/hey"
      }
    }
  }
}

Navigate to:

• /connect/github to request the public_repo scope
• /connect/github/notifications to request the notifications scope using another OAuth App (key and secret)
• /connect/github/all to request a bunch of scopes and also receive the response data in another callback route

Callback

Callback: Data

By default the response data will be returned in your callback route or absolute URL encoded as querystring.

Depending on the transport being used the response data can be returned in the session or in the state object instead.

The amount of the returned data can be controlled through the response configuration.

OAuth 2.0

{
  id_token: '...',
  access_token: '...',
  refresh_token: '...',
  raw: {
    id_token: '...',
    access_token: '...',
    refresh_token: '...',
    some: 'other data'
  }
}

The refresh_token is optional. The id_token is returned only for OpenID Connect providers requesting the openid scope.

OAuth 1.0a

{
  access_token: '...',
  access_secret: '...',
  raw: {
    oauth_token: '...',
    oauth_token_secret: '...',
    some: 'other data'
  }
}

Error

{
  error: {
    some: 'error data'
  }
}

Callback: Transport

querystring

By default Grant will encode the OAuth response data as querystring in your callback route or absolute URL:

{
  "github": {
    "callback": "https://site.com/hello"
  }
}

This is useful when using Grant as OAuth Proxy. However this final https://site.com/hello?access_token=... redirect can potentially leak private data in your server logs, especially when sitting behind a reverse proxy.

session

For local callback routes the session transport is recommended:

{
  "defaults": {
    "transport": "session"
  },
  "github": {
    "callback": "/hello"
  }
}

This will make the OAuth response data available in the session object instead:

req.session.grant.response // Fastify

state

The request/response lifecycle state can be used as well:

{
  "defaults": {
    "transport": "state"
  }
}

In this case a callback route is not needed, and it will be ignored if provided. The response data will be available in the request/response lifecycle state object instead:

res.grant.response // Fastify

Callback: Response

By default Grant returns all of the available tokens and the raw response data returned by the Authorization server:

{
  id_token: '...',
  access_token: '...',
  refresh_token: '...',
  raw: {
    id_token: '...',
    access_token: '...',
    refresh_token: '...',
    some: 'other data'
  }
}

querystring

When using the querystring transport it might be a good idea to limit the response data:

{
  "defaults": {
    "response": ["tokens"]
  }
}

This will return only the tokens available, without the raw response data.

This is useful when using Grant as OAuth Proxy. Encoding potentially large amounts of data as querystring can lead to incompatibility issues with some servers and browsers, and generally is considered a bad practice.

session

Using the session transport is generally safer, but it also depends on the implementation of your session store.

In case your session store encodes the entire session in a cookie, not just the session ID, some servers may reject the HTTP request because of HTTP headers size being too big.

{
  "google": {
    "response": ["tokens"]
  }
}

This will return only the tokens available, without the raw response data.

jwt

Grant can also return even larger response data by including the decoded JWT for OpenID Connect providers that return id_token:

{
  "google": {
    "response": ["tokens", "raw", "jwt"]
  }
}

This will make the decoded JWT available in the response data:

{
  id_token: '...',
  access_token: '...',
  refresh_token: '...',
  raw: {
    id_token: '...',
    access_token: '...',
    refresh_token: '...',
    some: 'other data'
  },
  jwt: {id_token: {header: {}, payload: {}, signature: '...'}}
}

Make sure you include all of the response keys that you want to be returned when configuring the response data explicitly.

profile

Outside of the regular OAuth flow, Grant can request the user profile as well:

{
  "google": {
    "response": ["tokens", "profile"]
  }
}

Additionaly a profile key will be available in the response data:

{
  access_token: '...',
  refresh_token: '...',
  profile: {some: 'user data'}
}

The profile key contains either the raw response data returned by the user profile endpoint or an error message.

Not all of the supported providers have their profile_url set, and some of them might require custom parameters. Usually the user profile endpoint is accessible only when certain scopes were requested.

Dynamic Configuration

Dynamic: Instance

Every Grant instance have a config property attached to it:

var grant = Grant(require('./config'))
console.log(grant.config)

You can use the config property to alter the Grant's behavior during runtime without having to restart your server.

This property contains the generated configuration used internally by Grant, and changes made to that configuration affects the entire Grant instance!

Dynamic: State

The request/response lifecycle state can be used to alter configuration on every request:

req.grant = {dynamic: {subdomain: 'usershop'}} // Fastify

This is useful in cases when you want to configure Grant dynamically with potentially sensitive data that you don't want to send over HTTP.

The request/response lifecycle state is not controlled by the dynamic configuration, meaning that you can override any configuration key.

Any allowed dynamic configuration key sent through HTTP GET/POST request will override the identical one set using a state override.

Dynamic: HTTP

The dynamic configuration allows certain configuration keys to be set dynamically over HTTP GET/POST request.

For example shopify requires your shop name to be embedded into the OAuth URLs, so it makes sense to allow the subdomain configuration key to be set dynamically:

{
  "shopify": {
    "dynamic": ["subdomain"]
  }
}

Then you can have a web form on your website allowing the user to specify the shop name:

<form action="/connect/shopify" method="POST" accept-charset="utf-8">
  <input type="text" name="subdomain" value="" />
  <button>Login</button>
</form>

When making a POST request to the /connect/:provider/:override? route you have to mount a form body parser middleware before mounting Grant:

// fastify
var parser = require('fastify-formbody')
.register(parser)
.register(grant(config))

Alternatively you can make a GET request to the /connect/:provider/:override? route:

https://awesome.com/connect/shopify?subdomain=usershop

Any dynamic configuration sent over HTTP GET/POST request overrides any other configuration.