Choosing an install route
Which install fits your store: the wizard, the npm packages, the pinned tag or the HTTP call.
On this page
Every route installs the same two parts: the browser part, which sees people and browser agents, and the server part, which sees the fetchers that never run JavaScript. The routes differ only in how the code reaches your site.
By stack
| Your site | Browser part | Server part |
|---|---|---|
| Next.js, Vite React, Express or Hono | npx parlox init adds both | Added by the wizard |
| Vue, Svelte, Angular, or any other site with a build step | @parlox/browser from npm, init() in the entry file | @parlox/server on a Node or edge server; the HTTP call from any other server that renders the pages; none when nginx or a static host serves the build (what that misses) |
| Remix, SvelteKit, Astro, Deno | @parlox/browser from npm | parloxFetch() from @parlox/server/fetch |
| Shopify theme | The pinned tag in theme.liquid | None: Shopify serves the pages. Add the checkout pixel |
| WordPress, plain HTML, any site without a build step | The pinned tag | The HTTP call from your server, if you control it |
| Google Tag Manager | The Parlox tag template | As for your stack |
| A static site on Vercel | @parlox/browser, or the tag | Vercel Routing Middleware |
| A static site anywhere else (Netlify, Cloudflare Pages, GitHub Pages, S3) | @parlox/browser, or the tag | None: there is no server to run it |
| PHP, Python, Ruby, Go, .NET | The tag, or @parlox/browser if the frontend is built | The HTTP call |
The wizard covers Next.js, Vite React, Express and Hono, and a repository that holds a frontend and a backend. For everything else, the steps by hand are short; each page shows them.
npm or the tag
Where your site has a build step, use the npm package. The tag is the same tracker, built from the same source, for sites that have no build step.
@parlox/browser from npm | The pinned tag | |
|---|---|---|
| Version | Pinned by your lockfile, reviewed like any dependency | Pinned by its URL and integrity hash |
| Where the code is served from | Your own build and your own domain | gateway.parlox.io |
| The recorder for agent visits | A separate chunk your bundler builds | A second file from the same release, pinned by a hash inside the tag |
| Consent mode | consent: "required" | data-consent="required" |
| Turn recording off | recordAgents: false | Not possible |
| Proves domain ownership by itself | No (use PARLOX_VERIFY_TOKEN or DNS) | Yes, when it is in your homepage's HTML |
| Content-Security-Policy | connect-src | script-src and connect-src |
A server part is optional, and worth it
Without a server part Parlox still sees people and browser agents. It does not see the assistants' fetchers (ChatGPT-User, Claude-User, Perplexity-User) or the search and training crawlers, because none of them run JavaScript. Those requests are the pages assistants read when someone asks them about what you sell.
Add it where your pages come from a server you control. A store whose pages are served by a platform (a Shopify theme, a hosted builder) or a static host without middleware can install the browser part only.
Sales events
Whatever the route, cart and checkout steps and the purchase of record are a separate, small addition: Cart and checkout events and Purchases from the server.