What is an API?

A program’s API, or application programming interface, is its defined set of endpoints that other programs can interact with — the contract that says what you can ask for, how to ask, and what you’ll get back. The point of an API is that you don’t need to know how the other program works inside. You only need to know the contract.

The term is broad. A software library’s function signatures are an API. So is the set of URLs a web service exposes. In analytics work, “API” almost always means the second kind: a web API you call over HTTP.

How a web API call works

You send an HTTP request to a URL and get a structured response back. The pieces:

• A method — GET to read, POST to create, PATCH or PUT to update, DELETE to remove.
• A path — the resource you want, like /v1/charges.
• Parameters — filters, date ranges, page size, usually in the query string.
• Headers — including authentication, typically an API key or a bearer token such as a JWT.
• A status code in the response — 200 for success, 401 for bad credentials, 429 when you’ve hit a rate limit.

Most modern web APIs follow REST conventions: resources get their own URLs, HTTP methods carry the meaning of the operation, and requests don’t depend on server-side session state. The response body is nearly always JSON:

{
  "object": "charge",
  "id": "ch_3Nx8",
  "amount": 4900,
  "currency": "usd",
  "created": 1714500000,
  "paid": true
}

You’ll also run into GraphQL, where you send one query describing exactly the fields you want, and older SOAP or XML-RPC APIs in enterprise systems.

APIs and analytics

Most of the tools your company runs on — payments, CRM, support, version control, ad platforms — hold their data behind an API rather than a database you can query. Getting that data into a place you can analyze means calling the API on a schedule, handling pagination and rate limits, and loading the results into a data warehouse. That’s the extract step of an ETL pipeline, and it’s how most of our integration guides get data flowing.

A few practical things to watch for: APIs paginate, so a “list” endpoint rarely returns everything at once; they rate-limit, so pipelines need backoff; and they version, so an endpoint that works today can be deprecated later. Timestamps and IDs also need careful handling — see data types for why casting matters once the data lands.

Metabase has its own API too, which you can use to automate administration or embed analytics in another application.

Related terms

• JSON
• JWT
• ETL
• Serialization

Further reading

• Working with the Metabase API
• Metabase API documentation

Put it to work

• Stripe — Integration
• GitHub — Integration
• Build a data pipeline — Integration