Application Programming Interface

An Application Programming Interface is an agreed handover point between two computer programs. One program makes a request, the other delivers a response back. To do this, both must first agree on fixed rules: which requests are allowed, how they are formulated, and what the response looks like. A weather service can, for example, offer: send me a place name, and you get the temperature back. A weather app on a phone then uses exactly this handover point and doesn’t need to operate any measuring stations itself. The German term for this is Programmierschnittstelle, but in everyday usage almost everyone just says API.

Why hardly any service stands alone anymore

Almost every app you use is made up of foreign building blocks. A delivery app shows a map from a map provider, processes payment through a payment service provider, and sends notifications via yet another service. All of this runs through interfaces. Without them, every company would have to build each of these functions itself, which would be practically impossible.

For companies, interfaces are therefore also a business model. Providers of AI models hardly sell software for installation anymore. Instead, they sell access to their interface and bill per request. Anyone who reads such figures in business news is often reading indirectly about API usage.

Conversely, this creates dependency. If a provider triples its prices or shuts down an interface, all products built on it collapse. This has happened several times in recent years, for example with social networks that suddenly only gave out their data at a steep price. Developers therefore half-jokingly speak of the risk of having built on someone else’s land.

Request, response, and the contract in between

The most common design on the internet works using the same technology as a web browser. The requesting program calls up an address, similar to a web address, and attaches its parameters. The server responds with a data packet, usually in a text format called JSON. JSON is a simple notation of names and values that machines can easily read, for example temperature together with the value 14.

Documentation is important here. It is, in a sense, the interface’s instruction manual and describes every possible request. In addition, you often need a key, a so-called API key. This is a long string of characters that identifies the caller and ensures that usage can be billed.

Almost all providers also limit the number of requests per minute. This limit is called a rate limit and protects the server from overload. A common misconception is that an interface is the same as a website. The difference: a website delivers text and images for humans, an interface delivers pure data for programs.

APIs in AI services and in the news

When you use a chatbot in your browser, you see a finished interface. Behind it, your text goes to the model via an API. Other companies can address the same interface directly and integrate the model into their own products. This is how the same language assistant ends up in a school portal, an application tool, and a translation app.

Interfaces are now also mandatory in the financial sector. Banks in the EU must give third-party providers access to account data, provided the customer consents. This is exactly why a budgeting app can display transactions from several banks together. Reports about new payment services are almost always based on such banking interfaces.

This isn’t purely theoretical for you either. Many public bodies offer free interfaces, for example for timetables, weather data, or election results. With just a few lines of code, you can build your own small project from this. Getting started usually works via the documentation that belongs to every serious interface.