AI Newsway
Back to glossary

Application Programming Interface

Cloud computing

An API, or Application Programming Interface, is a defined contract that lets one piece of software request data or actions from another without knowing how the other is built. The contract specifies the available operations, the shape of the input, the shape of the response, and the errors that can occur, so either side can be rewritten as long as the interface holds. Web APIs dominate today. REST maps operations onto HTTP verbs and URLs and remains the default; GraphQL lets a client request exactly the fields it needs in one round trip; gRPC uses binary protocol buffers for fast service-to-service calls inside a system. Webhooks invert the direction, with the provider calling your endpoint when an event occurs instead of you polling. APIs are how SaaS products integrate and how AI reaches applications, since OpenAI, Anthropic, and Google all ship model access as an API. Production concerns are authentication, rate limits, versioning, and idempotency for retried requests. The pitfall is breaking changes: removing a field or tightening validation silently breaks every consumer, which is why stable APIs version explicitly and deprecate on a published schedule rather than changing in place.