DEPRECATION NOTICE: As of djangorestframework 3.9, the plan is to phase out CoreAPI in favor of OpenAPI as the default schema representation. New projects should consider using OpenAPI rather than CoreAPI. See the DRF 3.9 release announcment for more details.

Core API is a format-independent Document Object Model for representing Web APIs.

It can be used to represent either Schema or Hypermedia responses, and allows you to interact with an API at the layer of an application interface, rather than a network interface.

Core API currently has implementations available for Core JSON, Open API/Swagger, HAL, and JSON Hyper-Schema.

There is a command line tool that you can use to interact with APIs exposing any of these formats, as well as a Python client library.

Using a Core API client is a more robust and meaningful way to interact with your API than constructing HTTP requests and decoding responses. The dynamic client library is always up to date with the API, and client code focuses solely on the interface being provided, rather than dealing with network details and encodings.

Core API can be used to interact with any API that exposes a supported Schema or Hypermedia format.

There are various examples of using Core API as a client against existing APIs.

Below are two hypermedia style services available for demonstrating Core API.

You can interact with these example services either directly through your browser, by installing the command-line client, or by using one of the client libraries.

• Notes - Create, update and delete items from a list of notes. http://notes.coreapi.org/
• Game - Find the treasure in 5 turns or less. http://game.coreapi.org/

Note that because Core API can be used to provide multiple different output formats, it can also be used to provide APIs that can be interacted with directly in the browser, by returning an HTML rendered format of the API response.

Let's try interacting with the "Game" service using the command line client.

In this example, you have have 5 guesses to try to find the position of the hidden treasure in a 3x3 grid.

First make sure to install Python, then...

$ pip install coreapi-cli  # Use Python's package manager `pip` to install the command-line client.
$ coreapi get http://game.coreapi.org/
<Home "http://game.coreapi.org/">
    new_game()
$ coreapi action new_game
<Game "http://game.coreapi.org/dee07521-6862-4744-95ec-812db90135bd">
    board: "...a
            ...b
            ...c
            123"
    description: "5 turns remaining."
    new_game()
    play([position])
$ coreapi action play --param position=b3
<Game "http://game.coreapi.org/dee07521-6862-4744-95ec-812db90135bd">
    board: "...a
            ..xb
            ...c
            123"
    description: "4 turns remaining."
    new_game()
    play([position])

There are three layers to the Core API specification.

The following is an brief overview of the document layer, demonstrating how the client interacts with a Core API interface.

Documents are the basic building blocks of Core API.

Documents are key-value pairs that contain the data and actions presented by the interface. Documents always have an associated URL, and should also have a title.

The top level element in any Core API interface is always a Document.

Let's take a look at a Core API document by using the command line client.

$ pip install coreapi-cli
$ coreapi get http://notes.coreapi.org/
<Notes "http://notes.coreapi.org/">
    notes: [
        <Note "http://notes.coreapi.org/123d4e35-cb09-40c3-98d3-d119e9079fca">
            complete: true
            description: "Do the weekly shopping"
            delete()
            edit([description], [complete]),
        <Note "http://notes.coreapi.org/0ace0b3b-9db2-4c10-989e-76c5c61265e7">
            complete: false
            description: "Fix the kitchen door"
            delete()
            edit([description], [complete])
    ]
    add_note(description)

We've got a document here that contains a couple of other nested documents. We can also see the actions and data that the interface exposes.

Links are the available points of interaction that the interface presents.

Links have an associated URL and action, and may accept a set of named parameters.

Core API also has a concept of in-place transformations. Links that are marked as in-place effect a partial transformation on the document, modifying or removing the nested document from the document tree. The 'put', 'patch' and 'delete' actions default to being in-place.

Let's return to the command line client, and take a look at calling some links. We'll start by removing an existing note:

$ coreapi action notes 0 delete
<Notes "http://notes.coreapi.org/">
    notes: [
        <Note "http://notes.coreapi.org/123d4e35-cb09-40c3-98d3-d119e9079fca">
            complete: true
            description: "Do the weekly shopping"
            delete()
            edit([description], [complete])
    ]
    add_note(description)

Let's remove the final remaining note.

$ coreapi action notes 0 delete
<Notes "http://notes.coreapi.org/">
    notes: []
    add_note(description)

There should now be no notes remaining.

Okay, let's create a new note. In this case we'll want to include a named parameter when acting on the link.

$ coreapi action add_note --param description="Email venue about conference dates"
<Notes "http://notes.coreapi.org/">
    notes: [
        <Note "http://notes.coreapi.org/e7785f34-2b74-41d2-ab3f-f754f688987c/">
            complete: false
            description: "Email venue about conference dates"
            delete()
            edit([description], [complete])
    ]
    add_note(description)

Finally we'll update the state of the note we've just created:

$ coreapi action notes 0 edit --param complete=true
<Notes "http://notes.coreapi.org/">
    notes: [
        <Note "http://notes.coreapi.org/e7785f34-2b74-41d2-ab3f-f754f688987c/">
            complete: true
            description: "Email venue about conference dates"
            delete()
            edit([description], [complete])
    ]
    add_note(description)

Data primitives are the set of basic datatypes that may be used to represent data in the interface.

Core API supports the same subset of data primitives as JSON. These are Object, Array, String, Integer, Number, true, false, and null.

$ coreapi show notes 0 description
"Email venue about conference dates"
$ coreapi show notes 0 complete
true

Following a link may result in an error. An error is a set of key-value pairs, which is used to represent any error information associated with the failed transition.

Encountering an error prevents any transition from taking place, and will normally be represented by an exception or other error status by the client library.

For example, in this case the server responds with an error when we fail to include a parameter:

$ coreapi action add_note
<Error: Invalid parameters>
    description: [
        "This field is required."
    ]

In this case the server responds with an error when we include an invalid parameter:

$ coreapi action add_note -p description='xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxxxxxxxx-xxx'
<Error: Invalid parameters>
description: [
    "Ensure this field has no more than 100 characters."
]

For news and updates follow @core-api, or @_tomchristie.

For discussion of the tools and specification, use the Hypermedia Web mailing list.