Skip to content
Module
Tech

What is an API? A practical guide for PMs

An API (application programming interface) is a defined way for one piece of software to request data or actions from another. On the web, most APIs work over HTTP: a client sends a request to an endpoint (a URL) with a method such as GET or POST, and receives a response with a status code and usually JSON data. The documentation, often an OpenAPI specification, is the contract both sides rely on.

Definition written by Alex Dionisio, Senior Product Manager. Updated November 13, 2026.

Why it matters for a PM

Integrations with payment providers, CRMs, AI models and partners all go through APIs, and their limits shape the product: rate limits, pagination, authentication, the errors they return and how old versions are retired. A PM who can read API documentation and send a test request scopes an integration accurately and writes a spec that covers the failure cases.

Example

A PM scoping a CRM integration reads the API documentation and notices that contacts can only be fetched 100 at a time, with a limit of 10 requests per second. Syncing 200,000 contacts will take time, so the spec plans an initial import in the background and incremental updates afterward.

How does an API call work?

In a REST API, the most common style on the web, the client sends a request to an address called an endpoint, with a method that signals the action: GET reads, POST creates, PUT or PATCH updates, DELETE removes. The request carries headers, including the authentication key or token, and often a body in JSON format. The server replies with a status code and, usually, a JSON body of its own.

The documentation describes, for each endpoint, the accepted parameters, the shape of the response and the possible errors. Many APIs publish that description in a standard format, OpenAPI, which can generate interactive documentation and lets you try calls without writing any code.

REST, GraphQL or webhooks: what are the differences?

REST organizes an API around resources (customers, orders, invoices), each with its own address. GraphQL exposes a single entry point where the client describes exactly which fields it wants, which avoids stacking calls to fill a rich screen, at the price of a more demanding server setup. gRPC, common between internal services, favors performance with a binary format. SOAP, an older XML-based style, still turns up in banking and public-sector systems.

A webhook reverses the direction: instead of polling an API at regular intervals to find out whether something changed, your system receives a call the moment the event happens, such as a confirmed payment. Most integrations combine the two: the webhook gives notice, the API supplies the details.

How do you assess a partner's API before committing?

Read the documentation with a few concrete questions in mind. Does the API cover everything the user journey requires, or is some data visible only in the partner's own interface? What rate and volume limits apply? Are there webhooks, so you do not have to poll in a loop? Is there a sandbox environment? How are changes announced, and how long does an old version stay available?

A quick test with a tool such as Postman, or a plain curl command, often tells you more than the marketing page: how clear the error messages are, how consistent the field names are, how fast the responses come back. Check the status page and incident history as well, since an unreliable partner becomes your own reliability problem.

What should you plan when your product exposes its own API?

A public API is a product in its own right, with its own users: the developers working for your customers and partners. Who are they, and which tasks do they want to automate? How will you authenticate them and limit their access? How will the API evolve without breaking their integrations? What documentation, support and adoption metrics will you provide? Every published endpoint becomes a commitment that is hard to take back. Treat breaking changes like any other launch, with advance notice, migration guides and a clear end date for the old version.

Key points

  • The frontend calls the backend through an API; the backend in turn calls third-party APIs.
  • Status codes signal the outcome: 2xx for success, 4xx for a problem with the request, 5xx for a server failure.
  • Authentication relies on API keys or tokens, such as OAuth 2.0 access tokens, with scopes that limit access.
  • Versioning and deprecation policies tell you how long an integration will keep working.

Common mistakes

  • Specifying only the success path and leaving error handling to chance.
  • Ignoring rate limits until a data sync fails in production.
  • Putting API keys in frontend code or in shared documents.

Go further with Module

The courses and lessons that cover this concept:

Guides on this topic