Guides 6 min read

API integration: how it works and how to connect to an API

What an API integration is, how to connect to an API step by step (keys, test environment, webhooks or polling) and what drives the cost, with a worked example.

by Tsunami Digital Team May 13, 2026 Updated October 16, 2026

An API integration is an automatic connection between two programs through their APIs, so one system sends or reads data from the other without anyone retyping it. Example: a web shop pushes every new order straight into accounting, and the warehouse gets a packing request.

Quick answer

An API (Application Programming Interface) is the agreed way one program asks another for data or an action.
Connecting to an API takes seven steps: documentation, key or OAuth, test environment, endpoints, webhook or polling, error handling and logging.
A webhook reports a change the moment it happens; polling asks “anything new?” every few minutes.
Cost depends on the number of systems, the quality of their APIs, data mapping and monitoring needs.

What an API is, in business terms

Think of an API as a service counter with clear rules. One program sends a request (for example “give me orders since 10:00”) and the other returns the data in an agreed format, usually JSON.

When two systems talk like this continuously, with nobody in between, that is an API integration. It is the base of any serious process automation: data is entered once and every system sees it.

How to connect to an API, step by step

Documentation. Start with the vendor’s official docs: what data the API exposes, its rate limits and how long each version is supported.
API key or OAuth. You need a key (token) or an OAuth login, where a user approves access. Keys live on the server, never in the browser or in an email.
Test environment. Connect to a test account or sandbox first, so trial orders never land in your real books.
Endpoints. Decide exactly which addresses and fields you need: orders, products, stock, customers. Request only what you use.
Webhook or polling. If the system supports webhooks, let it tell you about changes. Use polling when there are no webhooks, or as a periodic check that nothing was missed.
Error handling. Decide what happens when the other system does not answer: retry with a delay, a queue and an alert to a person. The same record must never be written twice.
Logging and monitoring. Every call and every error should leave a trace, and the alert must arrive before a customer reports the problem.

Worked example: a Shopify shop and accounting

Assumptions (illustrative): a shop on Shopify’s Standard plan, 60 orders a day, orders going into an accounting program. We checked the Shopify figures in the official docs on 16 October 2026.

What a request looks like. The integration sends a POST request to the shop address with the API version in the path (for example /admin/api/2026-10/graphql.json), and the key goes in the X-Shopify-Access-Token header. The response holds the requested orders plus how many points the query used and how many are left. Shopify has treated the old REST Admin API as legacy since 1 October 2024, so new integrations are built on GraphQL (Shopify).

The limit. The Standard plan allows 100 points per second, and a single query may not exceed 1,000 points (Shopify). Each write (mutation) costs 10 points by default.

Calls per day:

Method	Calls per day	Delay
Polling every 5 minutes	24 × 12 = 288	up to 5 minutes
Polling every minute	1,440	up to 1 minute
Webhook + hourly check	60 + 24 = 84	a few seconds

With 5-minute polling, at most 60 of the 288 queries bring a new order, so at least 228 (79 %) come back empty. A webhook with an hourly check needs 71 % fewer calls and reports the order within seconds.

Why keep the hourly check: Shopify states that webhook delivery is not guaranteed and recommends periodic reconciliation. Your server must answer within 5 seconds, and after a failure Shopify retries 8 times over 4 hours (Shopify).

Where the limit really bites: bulk changes. Updating the price of 5,000 products with one mutation each is 50,000 points, so at 100 points per second it takes at least about 8 minutes. Bulk operations exist for jobs like that.

What drives the cost of an API integration

We do not publish prices because they depend on scope. The first call is free, and the fixed price is agreed in writing before work starts. Effort depends mostly on:

The other side’s API. A modern API with docs and a sandbox is far simpler than a CSV export or an FTP drop, which is common with Croatian accounting software.
Direction of data. One-way order transfer is easier than two-way stock and price sync, as described in our guide to connecting a web shop to ERP.
Data mapping. Product codes, VAT rates, units and discounts must match in both systems.
Volume and speed. Real-time sync and thousands of records need queues and bulk operations.
Monitoring and maintenance. Vendors ship new API versions regularly (Shopify every three months) and retire old ones, so the integration needs a reliable development partner looking after it.

The price of CRM and ERP systems and their licences is a separate topic, covered in our article on CRM and ERP system cost. For large systems such as SAP and Pantheon, see ERP integrations.

Zapier and Make or a custom integration

Zapier and Make suit simple flows (one trigger, one action) and modest volume with ready-made connectors. They struggle with complex logic, high volume and tracing why a record got stuck.

A custom integration makes sense when the process is business-critical or no connector exists. We build these as part of our integrations service, and the wider workflow is covered by business process automation.

Frequently Asked Questions

What is an API integration? An API integration is an automatic connection between two programs through their APIs. One system sends or reads data from the other, so nothing is retyped by hand, for example web shop orders going straight into accounting.

How do you connect to an API? Read the documentation, get an API key or OAuth access, connect to a test environment first, pick the endpoints, choose between webhooks and polling, then add error handling and logging.

Which is better, a webhook or polling? A webhook is faster and uses fewer calls because the system reports a change as it happens. Polling is simpler and works where there are no webhooks. Most integrations use both.

How much does an API integration cost? It depends on the other side’s API, the direction and volume of data, mapping and monitoring. We do not publish prices, but after a free call we agree a fixed price in writing before work starts.

Who maintains the integration after launch? Under a maintenance contract, the development partner tracks new API versions and updates the code. Without one, the risk of an outdated version is yours.

Thinking about an API integration?

Tell us which systems you want to connect. On a free call we will check their APIs, suggest webhooks or polling and agree a fixed price before work starts.

All articles