b27544d Apr 18, 2017
@andyjohnson @garemoko @jhaag75 @Asoul @bryant1410 @oconnetf
661 lines (515 sloc) 36 KB

Experience API

Advanced Distributed Learning (ADL) Co-Laboratories


"Copyright 2013 Advanced Distributed Learning (ADL) Initiative, U.S. Department of Defense

Licensed under the Apache License, Version 2.0 (the "License"). You may not use this file except in compliance with the License. You may obtain a copy of the License at

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License."

This document was authored by members of the Experience API Working Group (see list in in support of the Office of the Deputy Assistant Secretary of Defense (Readiness) Advanced Distributed Learning (ADL) Initiative. Please send all feedback and inquiries to

Table of Contents

Part One: About the Experience API

1.0 Introduction

The Experience API (xAPI) is a technical specification that aims to facilitate the documentation and communication of learning experiences. It specifies a structure to describe learning experiences and defines how these descriptions can be exchanged electronically.

The xAPI is an effort of the Advanced Distributed Learning (ADL) Initiative. ADL was established in 1997 to standardize and modernize training and education management and delivery in the US. Since that time, there has been a growing realization of the need to track learning experiences of individuals beyond formal, structured computer-based training. In assessing candidates' suitability for positions or their capability for performing various tasks, there is a need to consider a wide range of formal and informal learning experiences, both on and offline. That information, more often than not is scattered across a wide variety of sources.

Out of this perceived need, the xAPI community and specification were born. xAPI assumes that:

  • There is a need to be able to analyze information about learning experiences and their outcomes distributed across a wide variety of sources, platforms and technologies.
  • Developing a commonly-accepted framework for gathering, storing and exchanging this information represents the best way of achieving this.

The goals of the xAPI are:

  • To make it easier to understand and compare learning experiences and their outcomes recorded across a wide variety of contexts, platforms and technologies.
  • To maximize interoperability of services which create, gather, store and process information about learning experiences.
  • To provide a guide to those who want to build applications that conform to and implement this specification.
  • To provide criteria against which conformance to this specification can be tested.

The document that follows sets out the xAPI specification which is designed to achieve these goals.

2.0 How to Use This Document

This is a definitive document which describes how the Experience API is to be implemented. It is a technical document authored specifically for individuals and organizations implementing this technology with the intent of such individuals developing interoperable tools, systems and services that are independent of each other and interoperable with each other.

Whenever possible, the language and formatting used in this document is intended to be considerate of non-technical readers because various tools, systems and services are based on the specification set described below. For this reason, sections that provide a high-level overview of a given facet of the Experience API are labeled description or rationale. Items in this document labeled as requirements, details or examples are more technical.

This specification is split into three parts. Part one is this introduction. It offers some background, high-level summaries and direction on how to read the rest of the specification.

Part two of this specification defines a data model for various data objects that are used in this specification. The most significant object within the xAPI data model is the "Statement" object. This specification defines the properties of the Statement object (including "Actor", "Verb", "Object", "Result", and "Context") and the rules of syntax for the values of those properties and how they are represented. This part helps to ensure that services implementing the specification follow a consistent data structure.

Part three of this specification sets out the transfer methods that must be used when communicating information about learning experiences between services that adhere to the specification. This includes the format of requests and the expected responses. Note that communication in xAPI is not restricted to a "Learning Record Store" (LRS) receiving data from "content". LRSs can communicate with services ranging from "Learning Record Providers" to "Learning Record Consumers" to other LRSs. xAPI follows the guidelines of the REST software architecture style, and as such data is transferred via HTTP requests and responses. Part three also defines security methods allowing for the trusted exchange of information between the LRS and trusted "Clients".


There are three levels of obligation with regards to conformance to the xAPI specification identified by the terms MUST, SHOULD and MAY. A service or system that fails to implement a MUST (or a MUST NOT) requirement is non-conformant. Failing to meet a SHOULD requirement is not a violation of conformity, but goes against the recommendations of the specification. MAY indicates an option, to be decided by the developer with no consequences for conformity. Usage of these terms outside of requirement language does not designate a requirement and is avoided whenever possible. Complete definitions of MUST, SHOULD, MAY, MUST NOT and SHOULD NOT are found in RFC 2119.

The use of an asterisk* following SHOULD indicates a very strong recommendation. It is planned that these recommendations will become MUST requirements in a future version. Not following these recommendations could risk interoperability and and/or lead to various other issues depending on the specifics of the recommendation. These recommendations cannot be MUST requirements within this version as these would be breaking changes. The xAPI Working Group strongly encourages adopters to implement these requirements as though they were MUST requirements, while continuing to support other adopters that might not do so.

2.2 Guidelines for Interpreting Descriptive Text and Tables

As a rule of thumb, if the guideline appears technical or seems to be a requirement, interpret it as such. This is especially true of longer, more, detailed explanations and of tables, each of which would be unintuitive and/or lengthy to dissect into a list of requirements.

Tables are used throughout this specification to define requirements for lists of properties, parameters, etc. The tables define which properties are required, recommended and optional. Generally, the notion of "optional" relates to the service creating the object, while services receiving and interpreting the object need to be able to interpret all properties of that object. Often, properties are optional because the data may not be relevant in every context; if the data is relevant in a particular context, then it is expected the property will be populated.

If an optional property or parameter contains an object with properties that are recommended or required, then these properties are only recommended/required if the property or parameter containing them is used.

Examples are provided throughout the specification and in appendices to illustrate implementation. The content of these examples is fictional in order to illustrate the requirements of the specification and may not always illustrate the best practice approach to tracking the particular learning experience used in the example. Examples can be used to inform interpretation of requirements, but are not intended to take precedence over requirements.

Where the specification does not include requirements relating to a particular facet of implementation, that detail can be considered to be outside of the scope of this specification. It is up to the implementer to determine a sensible approach. This specification tries to avoid vagueness and will usually give a rationale even if there no requirement in a given area.

3.0 Serialization and JavaScript Object Notation (JSON)

Serialization is the process of translating data objects and structures into a format for storage or transmission, such that the original data object can be recreated from the resulting serialization. In some cases it might be possible to serialize a piece of data in more than one way. For example, a boolean property with a value corresponding to true might be represented as true or 1 depending on the serialization used.

xAPI follows the rules of JSON for serializations (so boolean values are represented as true or false). It might also be possible to represent the objects defined in this specification using other serializations, such as XML. This is out of scope of this specification and use of anything other than JSON to represent the objects defined in this specification is not conformant.

Even within the rules of JSON there are possible variations of how data can be serialized, especially in relation to data about time. This is significant as a number of features of xAPI rely on systems being able to determine whether two Statements are equivalent. See Immutability and exceptions for more details about the properties of the Statement affected by this.

JSON allows for objects to have properties that contain empty objects. This is not recommended within xAPI; if the Statement is not intended to contain data about a property then it is expected that the property will not be used at all. All required properties are required to contain values.

4.0 Definitions

Activity: A type of Object making up the "this" in "I did this"; it is something with which an Actor interacted. It can be a unit of instruction, experience, or performance that is to be tracked in meaningful combination with a Verb. Interpretation of Activity is broad, meaning that Activities can even be tangible objects such as a chair (real or virtual). In the Statement "Anna tried a cake recipe", the recipe constitutes the Activity in terms of the xAPI Statement. Other examples of Activities include a book, an e-learning course, a hike, or a meeting.

Activity Provider (AP): Now referred to as a Learning Record Provider. This change differentiates that the activity itself is not always the responsibility of software, rather just the tracking portion is.

Actor: An individual or group representation tracked using Statements performing an action within an Activity. Is the "I" in "I did this".

Application Programming Interface (API): A set of rules and standards created to allow access into a software application or tool.

Authentication: The concept of verifying identity. Authentication allows interactions between two "trusted" parties.

Authorization: The affordance of permissions based on role; the process of making one party "trusted" by another.

Client: Refers to any entity that might interact through requests. Some examples could be a Learning Record Provider, a Learning Record Consumer, a Learning Record Store (LRS), or a Learning Management System (LMS).

Community of Practice (CoP): A group of practitioners connected by a common cause, role or purpose, which operates in a common modality. CoPs are focused on implementing xAPI within a specific knowledge domain or use case. CoPs or independent developers, can create domain-specific vocabularies, profiles, and recipes. These practices usually involve work around defining use cases and curating the various vocabulary terms, synonyms, and other related metadata that might be preferred within a CoP. They can also reuse existing vocabularies, profiles, and recipes already published by other CoPs or participants of the xAPI community.

Document Profile Resource: A construct where information about the learner or activity is kept, typically in name/document pairs that have meaning to an instructional system component. Not to be confused with Profile.

Endpoint: An entry point in a service-oriented-architecture. xAPI mirrors this approach with resources by defining the IRI from which communication takes place as an endpoint.

Experience API (xAPI): The collection of rules articulated in this document which determines how learning experiences are defined, formatted, and exchanged so that independent software programs can exchange and make use of this information.

Immutable: Adjective used to describe things which cannot be changed. With some exceptions, Statements in the xAPI are immutable. This ensures that when Statements are shared between LRSs, multiple copies of the Statement remain the same.

Internationalized Resource Identifier (IRI): A unique identifier which could be an IRL. Used to identify an object such as a verb, activity or activity type. Unlike URIs, IRIs can contain some characters outside of the ASCII character set in order to support international languages.

IRIs always include a scheme. This is not a requirement of this standard, but part of the definition of IRIs, per RFC 3987. What are sometimes called "relative IRIs" are not IRIs.

Internationalized Resource Locator (IRL): In the context of this document, an IRL is an IRI that when translated into a URI (per the IRI to URI rules), is a URL.

Inverse Functional Identifier: An identifier which is unique to a particular persona or Group.

Learning Experience: An event associated with learning. It is highly diverse as far as what it can be.
Examples include reading a book, taking an online course, going on a field trip, engaging in self-directed research, or receiving a certificate for a completed course.

Learning Management System (LMS): "A software package used to administer one or more courses to one or more learners. An LMS is typically a web-based system that allows learners to authenticate themselves, register for courses, complete courses and take assessments" (Learning Systems Architecture Lab definition). An LMS in this document is used as an example of how a user is identified as "trusted" within a system and able to access its Learning Experiences.

Learning Record: An account of a learning experience that is formatted according to the rules of xAPI. A Learning Record takes on many forms, including Statements, documents, and their parts. This definition is intended to be all-inclusive.

Learning Record Consumer (LRC): An xAPI Client that accesses data from Learning Record Store(s) with the intent of processing the data, including interpretation, analysis, translation, dissemination, and aggregation.

Learning Record Provider (LRP): An xAPI Client that sends data to Learning Record Store(s). Often, the Learning Record Provider will create Learning Records while monitoring a learner as a part of a Learning Experience.

Learning Record Store (LRS): A server (i.e. system capable of receiving and processing web requests) that is responsible for receiving, storing, and providing access to Learning Records.

Metadata Consumer: A person, organization, software program or other thing that seeks to determine the meaning represented by an IRI used within this specification and/or retrieves metadata about an IRI. An LRS might or might not be a metadata consumer.

Metadata Provider: A person, organization, software program or other thing that coins IRIs to be used within this specification and/or hosts metadata about an IRI.

Persona: A set of one or more representations which defines an Actor uniquely. Conceptually, this is like having a "home email" and a "work email". Both are the same person, but have different data, associations, etc.

Profile: A specific set of rules and documentation for implementing xAPI in a particular context. Profiles generally provide a particular vocabulary of terms, some created specifically for the profile, and some are referenced from other vocabularies. Sometimes a profile might provide multiple vocabularies for different situations, and sometimes someone might curate a vocabulary from multiple sources without creating a profile. Not to be confused with Document Profile Resource.

Registration: An instance of an Actor experiencing a particular Activity.

Representational State Transfer (REST): An architecture for designing networked web services. It relies on HTTP methods and uses current web best practices.

Service: A software component responsible for one or more aspects of the distributed learning process. An LMS typically combines many services to design a complete learning experience.

Statement: A data sctructure showing evidence for any sort of experience or event which is to be tracked in xAPI as a Learning Record. A set of several Statements, each representing an event in time, might be used to track complete details about a learning experience.

Tin Can API (TCAPI): The previous name of the API defined in this document, often used in informal references to the Experience API.

Verb: Is the action being done by the Actor within the Activity within a Statement. A Verb represents the "did" in "I did this".

Vocabulary: A list or collection of the terms that are used by a COP for labeling or categorizing information in a particular domain. The use of a vocabulary ensures that everyone is using the same word to mean the same thing. For more information on vocabularies, see the xAPI Vocabulary Companion Specification.

5.0 xAPI Components

This section explains and shows graphically how different pieces of xAPI can fit together.

Tracking Experiences
Figure 1: Data Flow in xAPI

Data Flow in xAPI

Figure 1 shows the tracking of learning experiences. A learner has a learning experience. This experience could take place in an online course, it could be on the job or it could be part of recreation. It really could be anything. This experience is tracked, on the learner’s behalf, by a trusted Learning Record Provider (LRP). The Learning Record Provider can also be responsible for the trusted relationship between the experience and the learner. This might even include launching content for the learner and managing digital rights associated with the content.

The Learning Record Provider creates Learning Records and sends them to one or more Learning Record Stores (LRSs). The LRS stores the Learning Records and makes them available to any authorized Client. A Learning Record Consumer (LRC) is a type of Client that accesses Learning Records and makes use of them.

Activity Data and Metadata
Figure 2: xAPI Activity Data and Metadata

xAPI Activity Data and Metadata

Understanding how a single Activity (as uniquely identified by its IRI) is defined and described is a key concept in xAPI. Figure 2 shows this process. An Activity (as a part of a Statement) has metadata properties that may be populated within the Statement itself. This is done in the Statement's Activity Definition. The id of each Activity is an IRI which also could have metadata located at the resolution location of the IRI. Any metadata where the IRI resolves is under the control of the Metadata Provider. The Metadata Provider is also responsible for making sure the IRI is permanent and resolves correctly.

Any metadata located at where the IRI resolves is the authoritative source of metadata, and could be used to populate the LRS's canonical version of the Activity's metadata (LRS's Activity Definition) as a preference to what it receives from Statements. A Metadata Consumer can access metadata via the IRI for the authoritative version or can query the Activities Resource for the canonical version.

Agent and Persona Management
Figure 3: xAPI Agent and Persona Management

xAPI Activity Data and Metadata

xAPI grants the framework for allowing selective access to one's personal data. This is done through the management of what are called personas. In xAPI, each persona represents the "I" in "I did this" and is logically the subject of the Statement. Each Agent or Group in xAPI corresponds to a persona. A learner sending Learning Records to an LRS could have multiple personas (Agents) associated with him or her.

In Figure 3, a learner accesses multiple services. Some of these services are used at work, others at home. Some are used for social purposes and others are used for educational or professional puproses. Thus, there are multiple personas functioning collectively within these services. Each of these services send data to the LRS. Afterwards, there are Statements from three different personas of the same learner.

The LRS can aggregate all of the information of each of the personas into one "Person" Object. This object can be retrieved from The LRS via the Agents Resource. How the LRS knows that these multiple personas belong to a single person is out of scope of this specification and there are several different approaches that an LRS can take. It's also possible that some LRSs will have no mechanism for associating personas.

6.0 Extending xAPI

xAPI can be extended in a few ways. The most notable are Statement extensions, which allow great flexibility within Statements. It is recommended that Communities of Practice agree on how to use extensions for their particular use cases and utilize profiles whenever possible. Implementation details are covered in 4.1 Extensions.

The About Resource is another instance where xAPI supports extensions. The LRS may find it useful to communicate features or behaviors beyond this specification. The LRS can use extensions to the About Resource to facilitate this communication.

Finally, the set of resources implemented is not expected to be constrained by this document. Resources beyond those listed in this specification can be implemented and co-exist with the resources defined in this specification.

7.0 Profiles, Vocabularies, and Communities of Practice

xAPI strictly defines the structure of Statements, but is very flexible as to the contents of that structure. For example, the specification requires that all Statements have a "verb" property, but does not restrict the values of that property; any Verb can be used. This flexibility enables xAPI to be used in any context, including future use cases not envisaged by the specification authors.

It is intended that Communities of Practice (CoPs) will define Verbs, Activity types, contextual relationships, extensions, etc. to be used in their profiles by providing unique identifiers wherever applicable. The CoP will define these identifiers and their metadata in a Vocabulary. A profile is a set of rules and vocabularies to implemented in addition to xAPI for the particular use case being addressed. It is very important that such communities exist and share best practices. For more information on publishing vocabularies, see the Vocabulary Companion Specification and Vocabulary Primer documents.

It is recommended that a profile use a unique "category" within a Statement's context to refer to any Statement
which implements the profile. An example profile is cmi5, which is designed for the traditional single learner, single online learning use case. An example cmi5 Statement can be found in Appendix B: cmi5 Example.

CoPs are highly recommended to avoid duplication of effort, as creating too many ways to solve the same problem will cause fragmentation in similar domains and can potentially hurt interoperability. An example of a CoP for the medical field is the MedBiquitous Learning Experience Working Group.


Appendix A: Revision History

0.8 (Project Tin Can API Deliverable) to 0.9 (March 31, 2012)

Rustici Software, who delivered the Project Tin Can API, made modifications to the API prior to the April 2012 Kickoff Meeting. It was voted in this meeting to move those changes into the current specification and revision to 0.9.

0.90 to 0.95 (August 31, 2012)

"Core" Verbs and Activity types were removed from the specification. References to these verbs in results, context, interactions, and Activity Definitions were also removed. It was recommended that implementers prefer community defined verbs to creating their own verbs.

  • Verbs, Activity types, and extension keys are now URIs
  • Restructured and added language around some of the other implementation details and scope.
  • Changed from using a person-centric view of Agents to a persona-centric view.
  • Friend of a Friend (FOAF) Agent merging requirement was removed.
  • Agent Objects now have exactly 1 uniquely identifying property, instead of at least one.
0.95 to 1.0.0 (April 26, 2013)

Various refinements and clarifications including:

  • Adding Attachments
  • Activity metadata is now stored as JSON rather than XML
  • Changes to voiding Statements
  • Clarification and naming of the Document APIs (later renamed to Document Resources)
  • Changes to querying the Statement API
  • Signed Statements


1.0.0 to 1.0.1 (October 1, 2013)

Clarifications and additional examples including:

  • Fixed various typos
  • Added additional examples in the appendices


1.0.1 to 1.0.2 (October 1, 2014)
  • Added optional/required to tables
  • Added missing table heading on Interaction components
  • Change yes/no to required/optional in attachments
  • Clarified intent of 'moreInfo' property


1.0.2 to 1.0.3 (September 21, 2016)
  • Complete reorganization of the document into three parts
  • Many, many clarifications
  • Definitions slightly modified and greatly clarified
  • Removed back-references to SCORM
  • Additional explanatory text and diagrams
  • Rename Statement API and Document APIs to Statement Resource and Document Resources


Appendix B: cmi5 Example

The following example illustrates a statement following the cmi5 Community of Practice. It demonstrates use of extensions and "category" contextActivities.

    "name":"Gert Frobe",