AngelGuard

Get started with AngelGuard

Monitor your AI models in production in five simple steps.

Step 1

Create your AngelGuard account

Go to https://dashboard.angelguard.com and create an account using your email address and a password.

You will receive a verification email. Once confirmed, you can access the AngelGuard dashboard.

After logging in, you can start configuring providers and models.

Step 2

Create a provider

From the dashboard, open the Providers menu and click New Provider.

A provider represents the source of the AI model you want to monitor (OpenAI, Azure, or an internal model).

Select Custom if the provider is not listed or if the model is self-hosted.

Step 3

Register your AI model

Go to the Models menu and click New Model.

Fill in the model configuration and select the provider created earlier.

Click Create Model to register it.

Step 4

Define detection rules and automation

Open the model details page and navigate to Rules & Automation.

Click Register New Rule to define anomaly detection conditions and automated actions.

Step 5

Connect your application

In the model details page, open the Access Keys section and copy your API key.

Send requests to your AI model through the AngelGuard proxy using this key.

Alerts and automated actions will trigger based on your configured rules.

const targetUrl = "https://api.openai.com/v1/responses";

const proxyUrl = new URL(
  "https://api.angelguard.com/api/model-proxy/proxy"
);
proxyUrl.searchParams.set("target", targetUrl);

const response = await fetch(proxyUrl.toString(), {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "X-AngelGuard-Key": process.env.ANGELGUARD_API_KEY,
    // Optional: forwarded to provider
    "Authorization": `Bearer ${process.env.PROVIDER_API_KEY}`,
  },
  body: JSON.stringify({
    prompt: {
      id: "support-chat",
      version: "v1",
    },
    input: [
      { role: "user", content: "Hello, can you help me?" },
    ],
  }),
});

const data = await response.json();
console.log(data);