Skip to main content
Identify a visitor, send a Request ID with your application action and verify the result on your server. This guide uses the supported ShieldLabs package API, not a standalone generated client.

Before you start

A modern browser, a registered HTTPS domain and a Public Key. Node.js is needed only for the package-manager setup. Register and verify the website domain in your analytics dashboard. Use credentials from the same domain and environment as the browser check. For a fresh account, start with Quick Start.

Get your credentials

Copy the Public Key for your domain from Integration > API keys. A public frontend environment variable may contain this key, but never a Private API Key or webhook signing secret.

Install the SDK

Use yarn add or pnpm add with the same package names if your project uses that package manager. Framework packages remain the application’s responsibility.

Add the integration

Avoid a top-level await load(...). When the agent is blocked, the rest of the module never runs, so the submit handler is never attached. Some build targets also reject top-level await. Call load() again wherever you need the agent rather than keeping its first promise: that promise stays rejected after a passing network error or a slow load, and a new call recovers from both. On your server, read the verdict for requestId with a ShieldLabs server SDK, for example identifications.get(requestId) in @shieldlabs-ai/node, which waits until the identification has been scored. The History row appears about 1 to 3 seconds after the browser call and can be refined for up to about 10 seconds while follow-up checks finish, so start the identification when the user begins the action, for example with identifyOnInteraction() (see Protect a form ).
Keep the page alive after identify() resolves. The agent posts the identification right after it hands over the request ID. Keep the page open until your own request has been sent, and do not navigate away the moment identify() resolves (for example with location.href = ... in its then callback): the agent’s post may not have gone out yet. For classic full-page form posts, start the identification early with identifyOnInteraction() (see the guide).
Test on a registered domain. ShieldLabs records identifications only for the domains registered in your account. On localhost the page still receives a requestId, but the identification is rejected with 401 and your backend never finds it. Test on a development domain with its own keys, as described in Environments.

Test the complete flow

  1. Run the application on the registered HTTPS domain with its Public Key.
  2. Trigger the protected form once and check that a Request ID is sent to your own backend.
  3. Retrieve that same ID using a server SDK or locate it in the analytics dashboard.
  4. Test a missing ID and an agent load failure: your backend must treat the action as unverified.
The /api/signup routes shown in examples belong to your application; they are not ShieldLabs API endpoints. Connect one of the server quick starts before testing the full action. A successful browser call does not prove scoring is complete.

Troubleshooting

  • No History row: confirm the registered domain, credential/environment match and that the browser remained open while collectors posted.
  • Missing or pending verdict: scoring is asynchronous. The server helper waits within a bounded budget; handle a missing result and API errors explicitly.
  • Authentication error: use the Private API Key for History, not a Public Key, Management Secret Key or MCP OAuth token.
  • Invalid webhook signature: verify the original raw bytes with the endpoint’s full signing secret, before trusting parsed JSON.

Next steps