What is an REST API?

REST (defined by Roy Fielding in 2000) organizes an API around resources rather than remote procedure calls: a URL like `/orders/123` represents a specific order, and the HTTP verb used against that URL determines the action — GET to read it, PUT to update it, DELETE to remove it. Responses typically use standard HTTP status codes (200 for success, 404 for not found, 429 for rate-limited) and a structured format like JSON, which makes REST APIs broadly interoperable across virtually any programming language or platform without custom tooling. RESTful design also emphasizes statelessness: each request should carry all the information needed to process it (typically via an auth token), rather than relying on server-side session state from a previous request, which makes REST APIs easier to scale horizontally behind a load balancer. This is part of why REST became the dominant pattern for both public APIs and internal service-to-service communication over the past two decades, even as alternatives like GraphQL and gRPC have carved out use cases where their tradeoffs fit better. For AI agent tooling specifically, most third-party integrations an agent calls — CRMs, calendars, ticketing systems — expose a REST API, which is why tool-calling frameworks are largely built around wrapping REST endpoints as callable functions.