Nutty Peanut Marshmallow
Share your code. npm Orgs help your team discover, share, and reuse code. Create a free org »


2.4.1 • Public • Published

Ketting - A hypermedia client for javascript


The Ketting library is an attempt at creating a 'generic' hypermedia client, it supports an opinionated set of modern features REST services might have.

The library supports HAL, Web Linking (HTTP Link Header) and HTML5 links. It uses the Fetch API and works both in the browsers and in node.js.


const ketting = new Ketting('');
// Follow a link with rel="author". This could be a HTML5 `<link>`, a
// HAL `_links` or a HTTP `Link:`.
const author = await ketting.follow('author');
// Grab the current state
const authorState = await author.get();
// Change the firstName property of the object. Note that this assumes JSON.
authorState.firstName = 'Evert';
// Save the new state
await author.put(authorState);


npm install ketting


yarn add ketting

Features overview

Ketting is a library that sits on top of a Fetch API to provide a RESTful interface and make it easier to follow REST best practices more strictly.

It provides some useful abstractions that make it easier to work with true hypermedia / HATEAOS servers. It currently parses HAL and has a deep understanding of links and embedded resources. There's also support for parsing and following links from HTML documents, and it understands the HTTP Link: header.

Using this library it becomes very easy to follow links from a single bookmark, and discover resources and features on the server.

Following links

One core tenet of building a good REST service, is that URIs should be discovered, not hardcoded in an application. It's for this reason that the emphasis in this library is not on URIs (like most libraries) but on relation-types (the rel) and links.

Generally when interacting with a REST service, you'll want to only hardcode a single URI (a bookmark) and discover all the other APIs from there on on.

For example, consider that there is a some API at This API has a link to an API for news articles (rel="articleCollection"), which has a link for creating a new article (rel="new"). When POSTing on that uri, the api returns 201 Created along with a Location header pointing to the new article. On this location, a new rel="author" appears automatically, pointing to the person that created the article.

This is how that interaction might look like:

const ketting = new Ketting('');
const createArticle = await ketting.follow('articleCollection').follow('new'); // chained follow
const newArticle = await{ title: 'Hello world' });
const author = await newArticle.follow('author');
// Output author information
console.log(await author.get());

Embedded resources

Embedded resources are a HAL feature. In situations when you are modeling a 'collection' of resources, in HAL you should generally just create links to all the items in the collection. However, if a client wants to fetch all these items, this can result in a lot of HTTP requests. HAL uses _embedded to work around this. Using _embedded a user can effectively tell the HAL client about the links in the collection and immediately send along the contents of those resources, thus avoiding the overhead.

Ketting understands _embedded and completely abstracts them away. If you use Ketting with a HAL server, you can therefore completely ignore them.

For example, given a collection resource with many resources that hal the relationshiptype item, you might use the following API:

const ketting = new Ketting('');
const articleCollection = await ketting.follow('articleCollection');
const items = await someCollection.followAll('item');
for (const item of items) {
   console.log(await item.get());

Given the last example, if the server did not use embedding, it will result in a HTTP GET request for every item in the collection.

If the server did use embedding, there will only be 1 GET request.

A major advantage of this, is that it allows a server to be upgradable. Hot paths might be optimized using embedding, and the client seamlessly adjusts to the new information.

Further reading:

Automatically parsing problem+json

If your server emits application/problem+json documents (RFC7807) on HTTP errors, the library will automatically extract the information from that object, and also provide a better exception message (if the title property is provided).

Node and Browser

Ketting works on any stable node.js version and modern browsers. To run Ketting in a browser, the following must be supported by a browser:

  • The Fetch API.
  • Promises (async/await is not required)



The 'Ketting' class is the main class you'll use to access anything else.


const options = {}; // options are optional
const ketting = new Ketting('', options);

2 keys or options are currently supported: auth and fetchInit.

auth can be used to specify authentication information. Supported authentication methods are:

  • HTTP Basic auth
  • OAuth2 Bearer tokens
  • OAuth2 managed client

Basic example:

const options = {
  auth: {
    type: 'basic',
    userName: 'foo',
    password: 'bar'

OAuth2 Bearer example:

const options = {
  auth: {
    type: 'bearer',
    token: 'bar'

OAuth2 Resource Owner Password Credentials Grant example:

const options = {
  auth: {
    type: 'oauth2',
    client: {
      clientId: 'fooClient',
      clientSecret: 'barSecret',
      accessTokenUri: '',
      scopes: ['test']
    owner: {
      userName: 'fooOwner',
      password: 'barPassword'

OAuth 2 Client Credentials Grant example:

const options = {
  auth: {
    type: 'oauth2',
    client: {
      clientId: 'fooClient',
      clientSecret: 'barSecret',
      accessTokenUri: '',
      scopes: ['test']

The fetchInit option is a default list of settings that's automatically passed to fetch(). This is especially useful in a browser, where there's a few more settings highly relevant to the security sandbox.

For example, to ensure that the browser automatically passed relevant cookies to the endpoint, you would specify this as such:

const options = {
  fetchInit : {
    credentials: 'include'

Other options that you may want to set might be mode or cache. See the documentation for the Request constructor for the full list.


Return a 'resource' object, based on it's url. If the url is not supplied, a resource will be returned pointing to the bookmark.

If a relative url is given, it will be resolved based on the bookmark uri.

const resource = client.getResource(''); // absolute uri
const resource = client.getResource('/foo'); // relative uri
const resource = client.getResource(); // bookmark

The resource is returned immediately, and not as a promise.


The follow function on the Ketting follows a link based on it's relation type from the bookmark resource.

const someResource = await ketting.follow('author');

This is just a shortcut to:

const someResource = await ketting.getResource().follow('author');


The fetch function is a wrapper for the new Fetch web standard. This function takes the same arguments (input and init), but it decorates the HTTP request with Authentication headers.

const response = await ketting.fetch('');


The Resource class is the most important object, and represents a REST resource. Functions such follow and getResource always return Resource objects.


Returns the current uri of the resource. This is a property, not a function and is always available.


A property representing the Content-Type of the resource. This value will be used in GET requests (with the Accept header) and PUT requests (with the Content-Type header).

The contentType might be available immediately if the current resource was followed from a link that had "type" information. If it's not available, it might be determined later, after the first GET request is done.


Returns the result of a GET request. This function returns a Promise.

await resource.get();

If the resource was fetched earlier, it will return a cached copy.


Updates the resource with a new representation

await resource.put({ foo: 'bar' });


Deletes the resource.

await resource.delete();

This function returns a Promise that resolves to null.

This function is meant to be an easy way to create new resources. It's not necessarily for any type of POST request, but it is really meant as a convenience method APIs that follow the typical pattern of using POST for creation.

If the HTTP response from the server was successful and contained a Location header, this method will resolve into a new Resource. For example, this might create a new resource and then get a list of links after creation:

const newResource = await{ foo: 'bar' });
// Output a list of links on the newly created resource
console.log(await newResource.links());


This function provides a really simply implementation of the PATCH method. All it does is encode the body to JSON and set the Content-Type to application/json. I'm curious to hear use-cases for this, so open a ticket if this doesn't cut it!

await resource.patch({
  foo: 'bar'


The refresh function behaves the same as the get() function, but it ignores the cache. It's equivalent to a user hitting the "refresh" button in a browser.

This function is useful to ditching the cache of a specific resource if the server state has changed.

console.log(await resource.refresh());


Returns a list of Link objects for the resource.

console.log(await resource.links());

You can also request only the links for a relation-type you are interested in:

resource.links('author'); // Get links with rel=author


Follows a link, by it's relation-type and returns a new resource for the target.

const author = await resource.follow('author');
console.log(await author.get());

The follow function returns a special kind of Promise that has a follow() function itself.

This makes it possible to chain follows:


Lastly, it's possible to follow RFC6570 templated links (templated URI), using the second argument.

For example, a link specified as:

{ href: "/foo{?a}", templated: true }

May be followed using

  .follow('some-templated-link', { a: 'bar' })

This would result following a link to the /foo?a=bar uri.


This method works like follow() but resolves into a list of resources. Multiple links with the same relation type can appear in resources; for example in collections.

var items = await resource.followAll('item');


The fetch function is a wrapper for the Fetch API. It takes very similar arguments to the regular fetch, but it does a few things special:

  1. The uri can be omitted completely. If it's omitted, the uri of the resource is used.
  2. If a uri is supplied and it's relative, it will be resolved with the uri of the resource.

For example, this is how you might do a HTTP PATCH request:

const init = {
  method: 'PATCH',
  body: JSON.serialize(['some', 'patch', 'object'])
const response = await resource.fetch(init);


This function is identical to fetch, except that it will throw a (async) exception if the server responded with a HTTP error.


The link class represents any Link of any type of document. It has the following properties:

  • rel - relation type
  • href - The uri
  • baseHref - the uri of the parent document. Used for resolving relative uris.
  • type - A mimetype, if specified
  • templated - If it's a URI Template. Most of the time this is false.
  • title - Human readable label for the uri
  • name - Unique identifier for the link within the document (rarely used).


Returns the absolute uri to the link. For example:

const link = new Link({ href: '/foo', baseHref: "" });
// output is


Expands a templated link. Example:

const link = new Link({ href: '{?q}', templated: true });
console.log(link.expand({ q: 'bla bla' });
// output is

OAuth2 Managed Client

The underlying OAuth2 client is implemented using js-client-oauth2 is exposed via the 'Ketting' class.

const ketting = new Ketting('', {
  auth: {
    type: 'oauth2',
    client: {
      clientId: 'fooClient',
      clientSecret: 'barSecret',
      accessTokenUri: '',
      scopes: ['test']
    owner: {
      userName: 'fooOwner',
      password: 'barPassword'
const oAuthClient = ketting.oauth2Helper.client;
// Interact with the underlying OAuth2 client


npm i ketting

Downloadsweekly downloads









last publish


  • avatar
Report a vulnerability