---
title: Scripts
description: Load vendor scripts, iframes and network requests in a SvelteKit
  app only after the visitor allows their consent category, and stop them when
  consent is withdrawn.
group: frameworks
lastModified: "2026-10-10T16:01:45+01:00"
---
## Register scripts in the root layout

Pass the array from `src/lib/scripts.ts` to `ConsentRoot`
as the `scripts` prop in `src/routes/+layout.svelte`. The
[quickstart](/docs/frameworks/sveltekit/quickstart) sets up both files:

```ts title="src/lib/scripts.ts"
import { posthog } from '@c15t/integrations/posthog';

export const scripts = [
	// Loads PostHog only after the visitor allows measurement.
	posthog({
		id: 'phc_your_project_key',
		initOptions: { cookieless_mode: 'never' },
		loadMode: 'after-consent',
	}),
];
```

Create the scripts in the layout component, not in `+layout.server.ts`. Script
configurations hold functions, which a server load cannot send to the
browser. Scripts load only in the browser, after hydration, so they never
appear in server HTML.

## Preload the script loader

The provider loads the script loader only on pages whose `scripts` array is
not empty or that have [network blocker](./network-blocker) rules. The loader
and the blocker share one separate chunk. Pages with neither never download
it. On a page with either, the browser would otherwise request the chunk after
the app's JavaScript has run, and a returning visitor's consented scripts and
held requests would wait for that extra request.

`consentManifest()` from `@c15t/svelte/vite`, already in `vite.config.ts`
from the quickstart, lets `c15tHandle` link the chunk from the page's
`<head>` with `<link rel="modulepreload">`. It does this whenever the
`sveltekit()` plugin is present; there is nothing else to add.

SvelteKit builds the server before the client, so the server cannot know the
chunk's file name. After the client build, `consentManifest()` writes the chunk URL
into the server output, before SvelteKit prerenders pages and before the
adapter copies the build. Then `c15tHandle` adds one link to every page whose
provider has scripts or blocker rules, prerendered pages included. The link
carries the provider's `nonce`, or else the nonce SvelteKit put on its own
scripts. It asks for low priority (`fetchpriority="low"`): the
runtime needs the chunk only after hydration, so the browser fetches your
app's own chunks first. It does nothing in `vite dev`.

Without the plugin, or without `c15tHandle` in `hooks.server.ts`, scripts
still load, one request later.

## How registered scripts load

The provider's `scripts` prop takes an array of script configurations. Each
has a category. The loader adds a script to the page when its category becomes
allowed and removes it when the category is withdrawn. Nothing optional loads
while the policy is still resolving, or when it fails.

Helpers in `@c15t/integrations`, such as `posthog()` from `@c15t/integrations/posthog`,
return a configuration with the right category and the vendor's own consent
calls. [Integrations](/docs/integrations/overview) lists every helper. For an
SDK without a helper, write a configuration with an `id`, `category` and `src`
as shown in [building integrations](/docs/integrations/building-integrations).

Remove the vendor's original `<script>` tag, `app.html` snippet or SDK import
before you register it. A banner does not block code that loads some other
way, and a vendor loaded twice sends events twice.

The `scripts` array is read when the provider is created. Build it once, at
the top level of the component, not inside an effect.

## Script options

Helpers set these for you. For a script without a helper, write the object
yourself:

|Option|Default|Behavior|
|--|--|--|
|`id`|required|A unique name. The loader uses it to add and remove the script once.|
|`category`|required|The category or condition, such as `'measurement'` or `{ and: ['measurement', 'marketing'] }`, that must be allowed.|
|`src` or `textContent`|none|The script's URL, or inline code.|
|`callbackOnly`|`false`|Adds no `<script>` element and only runs the callbacks. Use it to switch an SDK you load yourself on and off.|
|`alwaysLoad`|`false`|Loads at once, whatever the consent state. Only for scripts that apply consent themselves, such as a tag manager with Consent Mode.|
|`persistAfterConsentRevoked`|`false`|Keeps the element after withdrawal instead of removing it.|
|`target`|`'head'`|Where the element goes: `'head'` or `'body'`.|
|`async`, `defer`, `fetchPriority`, `attributes`, `nonce`|none|Set on the `<script>` element.|
|`anonymizeId`|`true`|Gives the element a random `id`, so ad blockers do not match it by name.|
|`vendor`|none|Also waits for this vendor to be allowed, for vendor-level consent outside IAB.|
|`onBeforeLoad`, `onLoad`, `onError`, `onConsentChange`, `onDispose`|none|Lifecycle callbacks. See [callbacks](./callbacks#script-callbacks).|

## Gate embeds

Iframes are not scripts. Wrap them in `ConsentGate`, which keeps the iframe
out of the DOM until its category is allowed:

```svelte title="src/YouTubeEmbed.svelte"
<script lang="ts">
	import { ConsentDialogLink, ConsentGate } from '@c15t/svelte';
</script>

<!-- The iframe mounts only while measurement is allowed. -->
<ConsentGate category="measurement">
	{#snippet placeholder()}<div class="placeholder">
			<p>Allow measurement to load this YouTube video.</p>
			<ConsentDialogLink>Choose video permissions</ConsentDialogLink>
		</div>{/snippet}
	<iframe
		title="YouTube video"
		src="https://www.youtube-nocookie.com/embed/czTksCF6X8Y?playsinline=1"
		allow="encrypted-media; picture-in-picture"
		allowfullscreen
	></iframe>
</ConsentGate>
```

For iframes from a CMS or Markdown, which you cannot wrap, use the iframe
blocker's `data-category` and `data-src` attributes. [Embeds](./embeds) covers
both.

## Block requests from code already on the page

The provider's `networkBlocker` option holds `fetch` and `XMLHttpRequest`
calls to the domains you list until their category is allowed. It is a
backstop for code you cannot move into `scripts`. [Network blocker](./network-blocker)
covers the rules and what it cannot stop.

## What happens when consent is withdrawn

Removing a script tag cannot stop code that already ran. So when a save
withdraws a category that was granted, the provider reloads the page after the
save request, and the new page starts with only the permitted code. Set
`reloadOnConsentRevoked: false` on the provider if you handle withdrawal
yourself, for example through a vendor's own opt-out call in
`onConsentChange`.

`ConsentGate` content unmounts without a reload. To delete first-party cookies
a vendor set, configure `clearOnRevocation`; see
[clearing data on revocation](./clear-on-revocation).

## Let visitors turn off one vendor

A visitor can allow marketing and still switch off one vendor in it. Pass
`vendors` to the provider; helpers from `@c15t/integrations` already carry their
vendor slug. See [vendor consent](./vendor-consent).

## Verify vendor loading

Open DevTools, clear site data for your origin and reload:

1. With the banner showing, the Network panel has no requests to your vendors,
   and gated iframes are absent from the Elements panel.
2. Allow one category in preferences. Only that category's vendors load, and
   its iframes appear.
3. Reject, reload, and confirm the vendor requests stay absent.
4. Withdraw a category you allowed. The page reloads and its vendors no longer
   load.

[Verify consent](/docs/guides/verify-consent) covers automated checks.
