AI API

Definitions

An application programming interface that gives software access to AI models as a hosted service, typically for text generation, embeddings, image generation or analysis, speech recognition, and speech synthesis. Instead of training or hosting a model, a developer sends an HTTP request containing input such as a prompt or an image and receives the model's output, often streamed token by token. Large language model APIs commonly bill per input and output token, enforce rate limits per minute, and offer features such as tool or function calling, structured output, and batch processing. Building on them introduces concerns that ordinary APIs rarely raise: outputs are non-deterministic, model versions are updated or retired on the provider's schedule, prompts may carry sensitive data subject to the provider's retention policy, and cost scales with usage in ways that are easy to underestimate. Many teams place a gateway or abstraction layer in front of several providers to manage these trade-offs.
In plain English: An API that gives applications access to AI features.
Example: "They used an AI API for summarization first so the product team could validate demand before investing in their own serving stack."