We publish our own traffic. If APIs.io is going to be the open place you go to discover 8329 providers, 28401 APIs, and 115763 schemas, then how much the directory actually gets used shouldn’t be a number we keep to ourselves. So it lives on a public page: apis.io/analytics, rebuilt every week straight from the raw numbers.
Here’s how to read it.
Two layers, two very different numbers
The thing to understand before anything else is that the page shows traffic measured at two different layers, and they don’t — and shouldn’t — agree.
- AWS CloudFront is the edge/CDN layer. It counts every request the CDN served — HTML pages, images and assets,
/apiand/mcpcalls, and every bot, crawler, and agent that came knocking. In the most recent week that was 5,363,220 edge requests moving 77.473 GB, at a 5.28% error rate. That error rate is edge-wide (mostly 4xx from bots probing old paths and missing assets), not a site-health alarm. - Google Analytics 4 is the browser/session layer. It only fires when a real browser loads a page and runs JavaScript, so it approximates humans: 161,481 page views, 152,684 sessions, and 149,272 users that same week.
Millions of edge requests, ~161K human page views. The gap between them isn’t an error — it’s the story. A huge share of what a modern API directory serves is machine traffic: crawlers indexing the catalog, agents grounding themselves in real API metadata through the MCP server, scripts hitting the REST API. CloudFront sees all of it; GA4 sees the people. Putting both on one page is the only honest way to show what “traffic” even means for a site like this.
Where the humans are
The GA4 card also breaks down the top countries for the week. Right now it’s led by the United States (77,459) and Singapore (39,308), followed by China (13,699), Germany (6,049), the Netherlands (6,046), Australia (3,089), the United Kingdom (1,209), and Japan (1,176) — a reminder that API discovery is a global, and heavily Asia-Pacific, activity.
Traffic over time
Below the weekly cards is a Chart.js line comparing CloudFront edge requests against GA page views across every week we’ve recorded — 27 of them now, going back to early January 2026. Both lines bend sharply upward through May and June, and the most recent week is the steepest yet. Plotting the edge layer and the human layer on the same timeline shows they’re growing together, not one at the expense of the other.
How it’s built
There’s no dashboard login and no client-side API call. A weekly GitHub Actions job pulls CloudFront metrics from CloudWatch and session data from the GA4 API, writes them into a single analytics.json, and Jekyll renders the page from that file at build time. The page you’re looking at is a static snapshot of real numbers, versioned in the repo like everything else on APIs.io.
That’s the whole idea. We ask providers to be transparent about their APIs — publish an OpenAPI, document your artifacts, show your work. The least we can do is publish ours. Go see the live version at apis.io/analytics.