我在免費 VPS 上運行 18 個正式環境 API，至今收入為 $0

I work shifts at a factory. On the side, I run 18 production APIs on a single free-tier cloud server, all of them listed on RapidAPI. Total paying subscribers so far: zero. This post is about how I got here, and what I got wrong.

The setup

The server is an Oracle Cloud Always Free ARM instance — 4 cores, 24 GB of RAM, $0 a month. Each API is a small FastAPI app on its own port, kept alive by a simple supervisor script that restarts it if it dies. RapidAPI sits in front as the gateway: it handles keys, quotas, and billing, so I don't have to build any of that myself.

The whole stack costs nothing to run. That was the point. I wanted to learn how real API products work without paying for the lesson.

What I built

The catalog grew in waves over about two weeks:

A background remover (my first one — it proved the pipeline end to end)
Utility APIs: screenshots, speech-to-text, text-to-speech, OCR, translation, IP geolocation, live prices
Validation APIs: email, phone, postal codes, domain health, VAT & IBAN, holidays
An AI email writer powered by free LLM tiers
A passive website security scanner — TLS, security headers, exposed-file checks, and a 0–100 score. That one is my flagship.

Every API has a free plan. The paid plans are cheap. The code runs on the VPS; the GitHub repo is the storefront with docs and examples.

The honest numbers

I checked my provider dashboard before writing this:

18 APIs live, 1 more (weather) still private
Paying subscribers: 0
Revenue: $0.00
API calls in the last 7 days: 23 — and every single one of them was me, testing my own endpoints during setup

The APIs work. I verified each one end to end through the real RapidAPI gateway, not just locally. The problem is that nobody knows they exist.

What I got wrong

I built first and thought about discovery later. Actually, I didn't think about discovery at all. Listing an API on a marketplace and waiting turns out to be a complete distribution strategy — a bad one.

Two weeks of building, zero weeks of telling anyone. The math is not subtle.

What I'm doing about it

I froze the catalog. No API #19 until people can actually find the first 18. The current plan:

Submit the APIs to directories and curated lists (one already found my repo on its own and asked to feature it — that email is what woke me up)
Write posts like this one, with real numbers, including the embarrassing ones
Keep the free tiers genuinely usable so people who do find the APIs get value fast