An application programming interface (API) is an interface or communication protocol between a client and a server intended to simplify the building of client-side software. It has been described as a “contract” between the client and the server, such that if the client makes a request in a specific format, it will always get a response in a specific format or initiate a defined action.^[1]

An API may be for a web-based system, operating system, database system, computer hardware, or software library.

An API specification can take many forms, but often includes specifications for routines, data structures, object classes, variables, or remote calls. POSIX, Windows API and ASPI are examples of different forms of APIs. Documentation for the API usually is provided to facilitate usage and implementation.

In building applications, an API simplifies programming by abstracting the underlying implementation and only exposing objects or actions the developer needs. While a graphical interface for an email client might provide a user with a button that performs all the steps for fetching and highlighting new emails, an API for file input/output might give the developer a function that copies a file from one location to another without requiring that the developer understand the file system operations occurring behind the scenes.^[2]

The term API seems to appear for the first time in the article of Ira W. Cotton, Data structures and techniques for remote computer graphics, published in 1968.

An API usually is related to a software library. The API describes and prescribes the "expected behavior" (a specification) while the library is an "actual implementation" of this set of rules.

A single API can have multiple implementations (or none, being abstract) in the form of different libraries that share the same programming interface.

The separation of the API from its implementation can allow programs written in one language to use a library written in another. For example, because Scala and Java compile to compatible bytecode, Scala developers can take advantage of any Java API.^[3]

API use can vary depending on the type of programming language involved. An API for a procedural language such as Lua could consist primarily of basic routines to execute code, manipulate data or handle errors while an API for an object-oriented language, such as Java, would provide a specification of classes and its class methods.^[4][5]

Language bindings are also APIs. By mapping the features and capabilities of one language to an interface implemented in another language, a language binding allows a library or service written in one language to be used when developing in another language.^[6] Tools such as SWIG and F2PY, a Fortran-to-Python interface generator, facilitate the creation of such interfaces.^[7]

An API can also be related to a software framework: a framework can be based on several libraries implementing several APIs, but unlike the normal use of an API, the access to the behavior built into the framework is mediated by extending its content with new classes plugged into the framework itself.

Moreover, the overall program flow of control can be out of the control of the caller and in the hands of the framework by inversion of control or a similar mechanism.^[8][9]

An API can specify the interface between an application and the operating system.^[10] POSIX, for example, specifies a set of common APIs that aim to enable an application written for a POSIX conformant operating system to be compiled for another POSIX conformant operating system.

Linux and Berkeley Software Distribution are examples of operating systems that implement the POSIX APIs.^[11]

Microsoft has shown a strong commitment to a backward-compatible API, particularly within its Windows API (Win32) library, so older applications may run on newer versions of Windows using an executable-specific setting called "Compatibility Mode".^[12]

An API differs from an application binary interface (ABI) in that an API is source code based while an ABI is binary based. For instance, POSIX provides APIs while the Linux Standard Base provides an ABI.^[13][14]

Remote APIs allow developers to manipulate remote resources through protocols, specific standards for communication that allow different technologies to work together, regardless of language or platform. For example, the Java Database Connectivity API allows developers to query many different types of databases with the same set of functions, while the Java remote method invocation API uses the Java Remote Method Protocol to allow invocation of functions that operate remotely, but appear local to the developer.^[15][16]

Therefore, remote APIs are useful in maintaining the object abstraction in object-oriented programming; a method call, executed locally on a proxy object, invokes the corresponding method on the remote object, using the remoting protocol, and acquires the result to be used locally as return value.

A modification on the proxy object also will result in a corresponding modification on the remote object.^[17]

Web APIs are the defined interfaces through which interactions happen between an enterprise and applications that use its assets, which also is a Service Level Agreement (SLA) to specify the functional provider and expose the service path or URL for its API users. An API approach is an architectural approach that revolves around providing a program interface to a set of services to different applications serving different types of consumers.^[18]

When used in the context of web development, an API is typically defined as a set of specifications, such as Hypertext Transfer Protocol (HTTP) request messages, along with a definition of the structure of response messages, usually in an Extensible Markup Language (XML) or JavaScript Object Notation (JSON) format. An example might be a shipping company API that can be added to an eCommerce-focused website to facilitate ordering shipping services and automatically include current shipping rates, without the site developer having to enter the shipper's rate table into a web database. While "web API" historically virtually has been synonymous for web service, the recent trend (so-called Web 2.0) has been moving away from Simple Object Access Protocol (SOAP) based web services and service-oriented architecture (SOA) towards more direct representational state transfer (REST) style web resources and resource-oriented architecture (ROA).^[19] Part of this trend is related to the Semantic Web movement toward Resource Description Framework (RDF), a concept to promote web-based ontology engineering technologies. Web APIs allow the combination of multiple APIs into new applications known as mashups.^[20] In the social media space, web APIs have allowed web communities to facilitate sharing content and data between communities and applications. In this way, content that is created in one place dynamically can be posted and updated to multiple locations on the web.^[21] For example, Twitter's REST API allows developers to access core Twitter data and the Search API provides methods for developers to interact with Twitter Search and trends data.

The design of an API has significant impact on its usage.^[2] The principle of information hiding describes the role of programming interfaces as enabling modular programming by hiding the implementation details of the modules so that users of modules need not understand the complexities inside the modules.^[22] Thus, the design of an API attempts to provide only the tools a user would expect.^[2] The design of programming interfaces represents an important part of software architecture, the organization of a complex piece of software.^[23]

Several authors have created recommendations for how to design APIs, such as Joshua Bloch,^[24] Kin Lane,^[25] and Michi Henning.^[26] Patterns for the design and evolution of remote APIs are covered in a series of EuroPLoP papers.^[27][28]

APIs are one of the more common ways technology companies integrate with each other. Those that provide and use APIs are considered as being members of a business ecosystem.^[29]

The main policies for releasing an API are:^[30]

An important factor when an API becomes public is its "interface stability". Changes by a developer to a part of it—for example adding new parameters to a function call—could break compatibility with the clients that depend on that API.^[32]

When parts of a publicly presented API are subject to change and thus not stable, such parts of a particular API should be documented explicitly as "unstable". For example, in the Google Guava library, the parts that are considered unstable, and that might change in the near future, are marked with the Java annotation @Beta.^[33]

A public API can sometimes declare parts of itself as deprecated or rescinded. This usually means that part of the API should be considered a candidate for being removed, or modified in a backward incompatible way. Therefore, these changes allows developers to transition away from parts of the API that will be removed or not supported in the future.^[34]

Client code may contain innovative or opportunistic usages that were not intended by the API designers. In other words, for a library with a significant user base, when an element becomes part of the public API, it may be used in diverse ways.^[35]

API documentation describes what services an API offers and how to use those services, aiming to cover everything a client would need to know for practical purposes.

Documentation is crucial for the development and maintenance of applications using the API.^[36] API documentation is traditionally found in documentation files but can also be found in social media such as blogs, forums, and Q&A websites.^[37]

Traditional documentation files are often presented via a documentation system, such as Javadoc or Pydoc, that has a consistent appearance and structure. However, the types of content included in the documentation differs from API to API.^[38]

In the interest of clarity, API documentation may include a description of classes and methods in the API as well as "typical usage scenarios, code snippets, design rationales, performance discussions, and contracts", but implementation details of the API services themselves are usually omitted.

Restrictions and limitations on how the API can be used are also covered by the documentation. For instance, documentation for an API function could note that its parameters cannot be null, that the function itself is not thread safe,^[39] or that a decrement and cancel protocol averts self-trading. Because API documentation tends to be comprehensive, it is a challenge for writers to keep the documentation updated and for users to read it carefully, potentially yielding bugs.^[32]

API documentation can be enriched with metadata information like Java annotations. This metadata can be used by the compiler, tools, and by the run-time environment to implement custom behaviors or custom handling.^[40]

It is possible to generate API documentation in data-driven manner. By observing a large number of programs that use a given API, it is possible to infer the typical usages, as well the required contracts and directives.^[41] Then, templates can be used to generate natural language from the mined data.

In 2010, Oracle Corporation sued Google for having distributed a new implementation of Java embedded in the Android operating system.^[42] Google had not acquired any permission to reproduce the Java API, although permission had been given to the similar OpenJDK project. Judge William Alsup ruled in the Oracle v. Google case that APIs cannot be copyrighted in the U.S, and that a victory for Oracle would have widely expanded copyright protection and allowed the copyrighting of simple software commands:

In 2014, however, Alsup's ruling was overturned on appeal to the Court of Appeals for the Federal Circuit, though the question of whether such use of APIs constitutes fair use was left unresolved.^[45]

In 2016, following a two-week trial, a jury determined that Google's reimplementation of the Java API constituted fair use, but Oracle vowed to appeal the decision.^[46] Oracle won on its appeal, with the Court of Appeals for the Federal Circuit ruling that Google's use of the APIs did not qualify for fair use.^[47] In 2019, Google appealed to the Supreme Court of the United States over both the copyrightability and fair use rulings.^[48]