SDKs and plugins

Five server-side SDKs, a browser loader, a WordPress plugin, a Tag Manager template and a Looker Studio connector.

The source for every one of them is in the product’s repository under ecosystem/, each with its own README and its own tests. If your package manager doesn’t have one yet, that’s where to get it.

The server-side SDKs

They exist for events that happen away from a browser: a payment webhook, a background job, an offline conversion, an API-only product with no pages at all.

All five are dependency-free. Nothing in this list pulls an HTTP library into your application.

// Go
client, _ := feasible.New(feasible.Options{Domain: "example.com"})
client.Pageview(ctx, feasible.FromRequest(r), "https://example.com/pricing")
client.Track(ctx, feasible.FromRequest(r), "Signup", "https://example.com/signup")

// Node
const analytics = createClient({ domain: "example.com" });
await analytics.pageview({ url: "https://example.com/pricing", ...visitorFromNodeRequest(req) });
await analytics.track("Signup", { url: "…", ...visitorFromNodeRequest(req) });

# Python
client  = Client(domain="example.com")
visitor = Visitor.from_request(headers=request.headers, remote_addr=request.remote_addr)
client.pageview(url="https://example.com/pricing", **visitor.as_kwargs())
client.event("Signup", url="…", **visitor.as_kwargs())

# Ruby
client  = Feasible.client(domain: "example.com")
visitor = Feasible::Visitor.from_request(request.env)
client.pageview(url: request.url, **visitor.to_h)

// PHP
$client  = new Client(domain: 'example.com');
$visitor = Visitor::fromRequest();
$client->pageview(...$visitor->args(), url: 'https://example.com/pricing');

The visitor’s address and user agent aren’t optional

Every one of these makes them required arguments rather than something you can forget.

Without them, every server-side event is geolocated to your data center and shares one visitor identifier - one visitor, from one country, forever - and there’s no way to repair that afterwards. Making it a compile error is the point.

Each SDK has a helper that pulls both off an incoming application request. Unlike the ingest service, an SDK helper has no trusted-proxy list; it assumes your application edge has already stripped client-supplied forwarding headers and written its own. It takes CF-Connecting-IP, then the first entry of X-Forwarded-For, then the socket address.

Don’t use that helper on a directly exposed app that accepts those headers from clients. Build the visitor from the socket address instead.

Attribution

An event with no browser has no referrer, so state it. referrer, utm_source, utm_medium, utm_campaign, utm_content and utm_term are all fields on the event. Left out, the visit is Direct. See the APIs.

Shared behavior

• Defaults: a five-second timeout and three attempts with exponential backoff. All of it is configurable in the constructor.
• A test mode that records instead of sending. Construct with disabled, or set FEASIBLE_DISABLED=1, and every event is kept in memory where your test can assert on it. An explicit false in code beats the environment variable, so a test suite can’t be silenced by a stray variable on a build machine.
• The drop reason comes back. Each call returns the result, including the x-feasible-dropped header if there was one, so a job can log why an event didn’t count instead of assuming it did.
• A debug call that returns the fully derived event without writing anything.

There are no mobile SDKs. iOS, Android and React Native are scoped and not built - post to the events endpoint directly.

@feasible/tracker - the browser loader

npm install @feasible/tracker

import { init, track } from "@feasible/tracker";
init({ domain: "example.com" });
await track("Signup", { props: { plan: "annual" } });

It loads the same script from the same host rather than bundling it, so you’re never running a stale copy.

What it adds is everything a hand-rolled tag gets wrong in a modern framework: it touches no browser global at module scope, so importing it in a server render is safe and does nothing; it installs the queue stub synchronously, so an event fired during hydration is replayed instead of throwing; init is idempotent, so a hot reload can’t double-count pageviews; and track always settles rather than hanging forever when a blocker eats the request.

enable(), disable() and isEnabled() drive the per-browser opt-out, and tell you whether the write happened.

WordPress

The plugin serves the script and the events endpoint from your own domain on randomized paths, which is what gets past ad blockers, and adds the site-search and 404 tracking a theme snippet can’t do.

Settings, proxy routes, path rotation and an embedded dashboard all live under one feasible.lol menu. See integrations.

Google Tag Manager

A tag template with two tag types - load the script, or send an event - and fields for the domain, host, hash routing, manual pageviews, counting localhost, a properties table and a revenue amount and currency.

It isn’t in Google’s gallery yet, so import template.tpl yourself under Templates → Tag Templates → New → Import.

The template declares permission to inject a script from one host only. If you proxy through your own domain, edit that permission in your own copy.

Looker Studio

A community connector you deploy into your own Apps Script project with clasp, then open through a datasources/create?connectorId=… link. It isn’t in Google’s gallery.

It authenticates with an API key carrying stats:read and sites:read. Self-hosters paste their host and key separated by a space.

It exposes 23 dimensions and 11 metrics, and pushes equality and contains filters down to us so the work happens here rather than in your browser. Results are cached for five minutes, thirty seconds for realtime, and errors are never cached.

A chart with two dimensions isn’t supported and says so, rather than returning a number that looks right and isn’t.