An API (application programming interface) is a defined doorway through which one program asks another to do something. For AI, it's how a model is reached by anything other than its own chat window: a developer's app, a workflow tool, or an agent sends a request to the model's API and gets an answer back. Almost every tool in this directory talks to a foundation model over an API.
You don't need to write code to care about this — the API is where the model stops being a chatbot and becomes a building block.
Why it matters at your desk. For a freelancer or small team, "is there an API?" decides whether you can wire an AI into your own process — your spreadsheet, your CRM, your scripts — rather than copy-pasting all day. API access is also usually billed per token, separately from a consumer subscription, so it's a different cost model to understand. The stakes show up in the news: new capabilities ship as APIs first (GPT-realtime voice), and API access is valuable enough to drive grey markets where it's officially restricted. Developer tools like Cursor live and die by the model APIs they connect to.
What to watch for: an API key is a credential that spends real money — leak it and someone else runs up your bill. Treat keys like passwords, and set spend limits where the provider allows.