Autorisation asynchrone

AI Identity for B2B permet aux agents d’IA d’autoriser les utilisateurs de manière asynchrone grâce au flux Client-Initiated Backchannel Authentication (CIBA). Les agents d’IA peuvent travailler en arrière-plan et n’avertir l’utilisateur qu’en cas de besoin, pour les actions critiques. Lorsque vous ajoutez des approbations sécurisées avec intervention humaine aux flux de production de vos agents d’IA, vous pouvez utiliser Auth0 pour demander à l’utilisateur son autorisation afin de mener à bien une requête d’autorisation. L’agent d’IA peut afficher des données d’autorisation enrichies dans l’invite de consentement, afin que l’utilisateur sache exactement ce qu’il autorise. À la fin de ce Démarrage rapide, vous disposerez d’un agent d’IA intégré au SDK Auth0 AI, capable de demander l’achat de produits dans une boutique en ligne au nom de l’utilisateur.

Choisissez votre pile technologique

LangGraph.js + Next.js
Vercel AI + Next.js
LangGraph + FastAPI
Vercel AI + Node.js

Download sample app

Start by downloading and extracting the sample app. Then open in your preferred IDE.

Install dependencies

In the root directory of your project, install the following dependencies:

@auth0/ai-langchain: Auth0 AI SDK for LangChain built for AI agents powered by LangChain.
@langchain/langgraph: For building stateful, multi-actor applications with LLMs.
langchain: The LangChain library.
@langchain/core: LangChain core libraries.
@langchain/openai: OpenAI provider for LangChain.
langgraph-nextjs-api-passthrough: API passthrough for LangGraph.

npm install @auth0/ai-langchain@4 @langchain/core@0.3.77 @langchain/langgraph@0.4.9 @langchain/openai@0.6.13 langchain@0.3.33 langgraph-nextjs-api-passthrough@0.1.4

Update the environment file

Copy the .env.example file to .env.local and update the variables with your Auth0 credentials. You can find your Auth0 domain, client ID, and client secret in the application you created in the Auth0 Dashboard.

Set up Human-in-the-Loop approvals

Integrate the Auth0 AI SDK into your application to secure your async AI agent workflow. For this quickstart, we will use a blocking request flow. In real use cases, often an asynchronous flow is preferred.

Configure the Auth0 AI SDK

To require asynchronous authorization for your tool, the tool needs to be wrapped with the Async authorizer, withAsyncAuthorization(). Let’s create a helper function to wrap the tool with the Async authorizer.Create a file at src/lib/auth0-ai.ts and instantiate a new Auth0 AI SDK client:

src/lib/auth0-ai.ts

import { Auth0AI } from "@auth0/ai-langchain";
import { AccessDeniedInterrupt } from "@auth0/ai/interrupts";

const auth0AI = new Auth0AI();

// Async Authorization flow for user confirmation
export const withAsyncAuthorization = auth0AI.withAsyncAuthorization({
  userID: async (_params, config) => {
    return config.configurable?.langgraph_auth_user?.sub;
  },
  bindingMessage: async ({ product, qty }) =>
    `Do you want to buy ${qty} ${product}`,
  scopes: ["openid", "product:buy"], // add any scopes you want to use with your API
  audience: process.env["SHOP_API_AUDIENCE"]!,

  /**
   * Controls how long the authorization request is valid for.
   */
  // requestedExpiry: 301,

  /**
   * The behavior when the authorization request is made.
   *
   * - `block`: The tool execution is blocked until the user completes the authorization.
   * - `interrupt`: The tool execution is interrupted until the user completes the authorization.
   * - a callback: Same as "block" but give access to the auth request and executing logic.
   *
   * Defaults to `interrupt`.
   *
   * When this flag is set to `block`, the execution of the tool awaits
   * until the user approves or rejects the request.
   * Given the asynchronous nature of the CIBA flow, this mode
   * is only useful during development.
   *
   * In practice, the process that is awaiting the user confirmation
   * could crash or timeout before the user approves the request.
   */
  onAuthorizationRequest: async (authReq, creds) => {
    console.log(`An authorization request was sent to your mobile device.`);
    await creds;
    console.log(`Thanks for approving the order.`);
  },

  onUnauthorized: async (e: Error) => {
    if (e instanceof AccessDeniedInterrupt) {
      return "The user has denied the request";
    }
    return e.message;
  },
});

This will intercept the tool call to initiate a Async Authorization request:

The Async Authorization request includes the user ID that will approve the request.
Auth0 sends the user a mobile push notification. The AI agent polls the /token endpoint for a user response.
The mobile application retrieves the bindingMessage containing the consent details, in this case, the details of the product to purchase.
The user responds to the request:
- If the request is approved, the tool execution will continue.
- If the request is rejected, the tool execution will not continue.

Pass credentials to the tools

Next, add the following code to src/lib/auth0.ts:

src/lib/auth0.ts

//... existing code
export const auth0 = new Auth0Client({
  authorizationParameters: {
    // In v4, the AUTH0_SCOPE and AUTH0_AUDIENCE environment variables are no longer automatically picked up by the SDK.
    // Instead, we need to provide the values explicitly.
    scope: process.env.AUTH0_SCOPE,
    audience: process.env.AUTH0_AUDIENCE,
  },
});

// Get the Access token from Auth0 session
export const getAccessToken = async () => {
  const tokenResult = await auth0.getAccessToken();

  if(!tokenResult || !tokenResult.token) {
    throw new Error("No access token found in Auth0 session");
  }

  return tokenResult.token;
};

Update the /src/app/api/chat/[..._path]/route.ts file with the following code. The access token will be passed to your LangGraph agent so we can use it from the Auth0 AI SDK to get the current user.

src/app/api/chat/[..._path]/route.ts

import { initApiPassthrough } from "langgraph-nextjs-api-passthrough";

import { getAccessToken } from "@/lib/auth0";

export const { GET, POST, PUT, PATCH, DELETE, OPTIONS, runtime } =
  initApiPassthrough({
    apiUrl: process.env.LANGGRAPH_API_URL,
    baseRoute: "chat/",
    headers: async () => {
      const accessToken = await getAccessToken();
      return {
        Authorization: `Bearer ${accessToken}`,
      };
  });

Add Custom Authentication

For more information on how to add custom authentication for your LangGraph Platform application, read the Custom Auth guide.

In your langgraph.json, add the path to your auth file:

langgraph.json

{
  "node_version": "20",
  "graphs": {
    "agent": "./src/lib/agent.ts:agent"
  },
  "env": ".env",
  "auth": {
    "path": "./src/lib/auth.ts:authHandler"
  }
}

Then, in your auth.ts file, add your auth logic:

src/lib/auth.ts

import { createRemoteJWKSet, jwtVerify } from "jose";

const { Auth, HTTPException } = require("@langchain/langgraph-sdk/auth");

const AUTH0_DOMAIN = process.env.AUTH0_DOMAIN;
const AUTH0_AUDIENCE = process.env.AUTH0_AUDIENCE;

// JWKS endpoint for Auth0
const JWKS = createRemoteJWKSet(
  new URL(`https://${AUTH0_DOMAIN}/.well-known/jwks.json`)
);

// Create the Auth instance
const auth = new Auth();
// Register the authentication handler
auth.authenticate(async (request: Request) => {
  const authHeader = request.headers.get("Authorization");
  const xApiKeyHeader = request.headers.get("x-api-key");
    /**
     * LangGraph Platform will convert the `Authorization` header from the client to an `x-api-key` header automatically
     * as of now: https://docs.langchain.com/langgraph-platform/custom-auth
     *
     * We can still leverage the `Authorization` header when served in other infrastructure w/ langgraph-cli
     * or when running locally.
     */
    // This header is required in Langgraph Cloud.
    if (!authHeader && !xApiKeyHeader) {
      throw new HTTPException(401, {
        message: "Invalid auth header provided.",
      });
    }

    // prefer the xApiKeyHeader first
    let token = xApiKeyHeader || authHeader;

    // Remove "Bearer " prefix if present
    if (token && token.startsWith("Bearer ")) {
      token = token.substring(7);
    }

    // Validate Auth0 Access Token using common JWKS endpoint
    if (!token) {
      throw new HTTPException(401, {
        message:
          "Authorization header format must be of the form: Bearer <token>",
      });
    }

    if (token) {
      try {
        // Verify the JWT using Auth0 JWKS
        const { payload } = await jwtVerify(token, JWKS, {
          issuer: `https://${AUTH0_DOMAIN}/`,
          audience: AUTH0_AUDIENCE,
        });

        console.log("✅ Auth0 JWT payload resolved!", payload);

        // Return the verified payload - this becomes available in graph nodes
        return {
          identity: payload.sub!,
          email: payload.email as string,
          permissions:
            typeof payload.scope === "string" ? payload.scope.split(" ") : [],
          auth_type: "auth0",
          // include the access token for use with Auth0 Token Vault exchanges by tools
          getRawAccessToken: () => token,
          // Add any other claims you need
          ...payload,
        };
      } catch (jwtError) {
        console.log(
          "Auth0 JWT validation failed:",
          jwtError instanceof Error ? jwtError.message : "Unknown error"
        );
        throw new HTTPException(401, {
          message: "Invalid Authorization token provided.",
        });
      }
    }
});

export { auth as authHandler };

Create a tool to call your API

In this example, we use a tool that buys products on the user’s behalf. When the user approves the transaction, the Auth0 AI SDK retrieves an access token to call the shop’s API. Upon completing the Async Authorization flow, the AI agent responds with a message confirming the purchase. The Auth0 AI SDK returns an error response if the user denies the transaction.Now, create a file src/lib/tools/shop-online.ts and add the following code:

src/lib/tools/shop-online.ts

import { tool } from "@langchain/core/tools";
import { z } from "zod";

import { getAsyncAuthorizationCredentials } from "@auth0/ai-langchain";

export const shopOnlineTool = tool(
  async ({ product, qty, priceLimit }) => {
    console.log(`Ordering ${qty} ${product} with price limit ${priceLimit}`);

    const apiUrl = process.env["SHOP_API_URL"]!;

    if (!apiUrl) {
      // No API set, mock a response
      return `Ordered ${qty} ${product}`;
    }

    const headers = {
      "Content-Type": "application/json",
      Authorization: "",
    };
    const body = {
      product,
      qty,
      priceLimit,
    };

    const credentials = getAsyncAuthorizationCredentials();
    const accessToken = credentials?.accessToken;

    if (accessToken) {
      headers["Authorization"] = "Bearer " + accessToken;
    }

    const response = await fetch(apiUrl, {
      method: "POST",
      headers: headers,
      body: JSON.stringify(body),
    });

    return response.statusText;
  },
  {
    name: "shop_online",
    description: "Tool to buy products online",
    schema: z.object({
      product: z.string(),
      qty: z.number(),
      priceLimit: z.number().optional(),
    }),
  }
);

Update environment variables

You need to obtain an API Key from OpenAI or another provider to use an LLM.If you want to use an API, it must be registered with Auth0 and have a valid audience.Update the .env.local file with the following variables:

.env.local

# ... existing variables
# You can use any provider of your choice supported by Vercel AI
OPENAI_API_KEY="YOUR_API_KEY"

# API
SHOP_API_URL=<your-shop-api-url>
SHOP_API_AUDIENCE=sample-shop-api

Require async authorization for your tool

Call the tool from your AI app to make purchases. Update the src/lib/agent.ts file with the following code:

src/lib/agent.ts

//...
import { withAsyncAuthorization } from "./auth0-ai";
import { shopOnlineTool } from "./tools/shop-online";

//... existing code

const tools = [
  //... existing tools
  withAsyncAuthorization(shopOnlineTool),
];

//... existing code

Test the application

Start the application with npm run all:dev. Then, navigate to http://localhost:3000.

This will open the LangGraph Studio in a new tab. You can close it as we won’t require it for testing the application.

You can ask the AI agent to buy a product, for example, “Buy an XYZ phone.” Now, look for a push notification from the Auth0 Guardian app or your custom app integrated with the Auth0 Guardian SDK on your mobile device. Once you approve the notification, you should see the tool being executed and a response from the agent.Explore the example app on GitHub.

Download sample app

Start by downloading and extracting the sample app. Then open in your preferred IDE.

Install dependencies

In the root directory of your project, install the following dependencies:

@auth0/ai-vercel: Auth0 AI SDK for Vercel AI built for AI agents powered by the Vercel AI SDK.
ai: Core Vercel AI SDK module that interacts with various AI model providers.
@ai-sdk/openai: OpenAI provider for the Vercel AI SDK.
@ai-sdk/react: React UI components for the Vercel AI SDK.
zod: TypeScript-first schema validation library.

npm install @auth0/ai-vercel@5 ai@6 @ai-sdk/openai@3 @ai-sdk/react@3 zod@3.25.76

Update the environment file

Set up Human-in-the-Loop approvals

Configure the Auth0 AI SDK

src/lib/auth0-ai.ts

import { Auth0AI } from "@auth0/ai-vercel";
import { AccessDeniedInterrupt } from "@auth0/ai/interrupts";

import { getUser } from "./auth0";

const auth0AI = new Auth0AI();

// Async Authorization flow for user confirmation
export const withAsyncAuthorization = auth0AI.withAsyncAuthorization({
  userID: async () => {
    const user = await getUser();
    return user?.sub as string;
  },
  bindingMessage: async ({ product, qty }) =>
    `Do you want to buy ${qty} ${product}`,
  scopes: ["openid", "product:buy"], // add any scopes you want to use with your API
  audience: process.env["SHOP_API_AUDIENCE"]!,

  /**
   * Controls how long the authorization request is valid for.
   */
  // requestedExpiry: 301,

  /**
   * The behavior when the authorization request is made.
   *
   * - `block`: The tool execution is blocked until the user completes the authorization.
   * - `interrupt`: The tool execution is interrupted until the user completes the authorization.
   * - a callback: Same as "block" but give access to the auth request and executing logic.
   *
   * Defaults to `interrupt`.
   *
   * When this flag is set to `block`, the execution of the tool awaits
   * until the user approves or rejects the request.
   * Given the asynchronous nature of the CIBA flow, this mode
   * is only useful during development.
   *
   * In practice, the process that is awaiting the user confirmation
   * could crash or timeout before the user approves the request.
   */
  onAuthorizationRequest: async (authReq, creds) => {
    console.log(`An authorization request was sent to your mobile device.`);
    await creds;
    console.log(`Thanks for approving the order.`);
  },

  onUnauthorized: async (e: Error) => {
    if (e instanceof AccessDeniedInterrupt) {
      return "The user has denied the request";
    }
    return e.message;
  },
});

This will intercept the tool call to initiate a CIBA request:

The CIBA request includes the user ID that will approve the request.
Auth0 sends the user a mobile push notification. The AI agent polls the /token endpoint for a user response.
The mobile application retrieves the bindingMessage containing the consent details, in this case, the details of the product to purchase.
The user responds to the request:
- If the request is approved, the tool execution will continue.
- If the request is rejected, the tool execution will not continue.

Next, add the following code to src/lib/auth0.ts:

src/lib/auth0.ts

//... existing code
export const getUser = async () => {
  const session = await auth0.getSession();
  return session?.user;
};

Require async authorization for your tool

In this example, we wrap a tool that buys products on the user’s behalf. When the user approves the transaction, the Auth0 AI SDK retrieves an access token to call the shop’s API. Upon completing the CIBA flow, the AI agent responds with a message confirming the purchase.The Auth0 AI SDK returns an error response if the user denies the transaction.Now, create and wrap the tool with the Async authorizer. Create a file src/lib/tools/shop-online.ts and add the following code:

src/lib/tools/shop-online.ts

import { tool } from "ai";
import { z } from "zod";

import { getAsyncAuthorizationCredentials } from "@auth0/ai-vercel";
import { withAsyncAuthorization } from "../auth0-ai";

export const shopOnlineTool = withAsyncAuthorization(
  tool({
    description: "Tool to buy products online",
    parameters: z.object({
      product: z.string(),
      qty: z.number(),
      priceLimit: z.number().optional(),
    }),
    execute: async ({ product, qty, priceLimit }) => {
      console.log(`Ordering ${qty} ${product} with price limit ${priceLimit}`);

      const apiUrl = process.env["SHOP_API_URL"]!;

      if (!apiUrl) {
        // No API set, mock a response or return error
        return `Ordered ${qty} ${product}`;
      }

      const headers = {
        "Content-Type": "application/json",
        Authorization: "",
      };
      const body = {
        product,
        qty,
        priceLimit,
      };

      const credentials = getAsyncAuthorizationCredentials();
      const accessToken = credentials?.accessToken;

      if (accessToken) {
        headers["Authorization"] = "Bearer " + accessToken;
      }

      const response = await fetch(apiUrl, {
        method: "POST",
        headers: headers,
        body: JSON.stringify(body),
      });

      return response.statusText;
    },
  })
);

Update environment variables

.env.local

# ... existing variables
# You can use any provider of your choice supported by Vercel AI
OPENAI_API_KEY="YOUR_API_KEY"

# API (OPTIONAL)
SHOP_API_URL=<your-shop-api-url>
SHOP_API_AUDIENCE=sample-shop-api

Integrate the tool into an AI agent

Call the tool from your AI app to make purchases. Update the src/app/api/chat/route.ts file with the following code:

src/app/api/chat/route.ts

//...
import { setAIContext } from "@auth0/ai-vercel";
import { shopOnlineTool } from "@/lib/tools/shop-online";

//...
export async function POST(req: NextRequest) {
  const { id, messages }: { id: string; messages: Array<UIMessage> } = await req.json();

  const tools = { shopOnlineTool };

  setAIContext({ threadID: id });

  const stream = createUIMessageStream({
    async execute({ writer }) {
      const result = streamText({
        model: openai.chat('gpt-4o-mini'),
        system: AGENT_SYSTEM_TEMPLATE,
        messages: await convertToModelMessages(messages),
        tools,
      });

      writer.merge(
        result.toUIMessageStream({
          sendReasoning: true,
        }),
      );
    },
  });

  return createUIMessageStreamResponse({ stream });
}
//...

Test the application

Start the application with npm run dev. Then, navigate to http://localhost:3000.You can ask the AI agent to buy a product, for example, “Buy an XYZ phone.” Now, look for a push notification from the Auth0 Guardian app or your custom app integrated with the Auth0 Guardian SDK on your mobile device. Once you approve the notification, you should see the tool being executed and a response from the agent.Explore the example app on GitHub.

Download sample app

Start by downloading and extracting the sample app. Then open in your preferred IDE.The project is divided into two parts:

backend/: contains the backend code for the Web app and API written in Python using FastAPI and the LangGraph agent.
frontend/: contains the frontend code for the Web app written in React as a Vite SPA.

Install dependencies

In the backend directory of your project, install the following dependencies:

auth0-ai-langchain: Auth0 AI SDK for LangChain built for AI agents powered by LangChain.
langgraph: LangGraph for building stateful, multi-factor applications with LLMs.
langchain-openai: LangChain integrations for OpenAI.
langgraph-cli: LangGraph CLI for running a local LangGraph server.

Make sure you have uv installed and run the following command to install the dependencies:

cd backend
uv sync
uv add "auth0-ai-langchain>=1.0.1" "langgraph>=0.5.4" langchain-openai "langgraph-cli[inmem]>=0.3.6" --prerelease=allow

Update the environment file

Copy the .env.example file to .env and update the variables with your Auth0 credentials. You can find your Auth0 domain, client ID, and client secret in the application you created in the Auth0 Dashboard.

Set up Human-in-the-Loop approvals

Configure the Auth0 AI SDK

To require asynchronous authorization for your tool, the tool needs to be wrapped with the Async authorizer, with_async_authorization(). Let’s create a helper function to wrap the tool with the Async authorizer.Create a file at app/core/auth0_ai.py and instantiate a new Auth0 AI SDK client:

app/core/auth0_ai.py

from auth0_ai.authorizers.types import Auth0ClientParams
from auth0_ai_langchain.auth0_ai import Auth0AI
from langchain_core.runnables import ensure_config

from app.core.config import settings

auth0_ai = Auth0AI(
    Auth0ClientParams(
        {
            "domain": settings.AUTH0_DOMAIN,
            "client_id": settings.AUTH0_CLIENT_ID,
            "client_secret": settings.AUTH0_CLIENT_SECRET,
        }
    )
)

with_async_authorization = auth0_ai.with_async_authorization(
    audience=settings.SHOP_API_AUDIENCE,
    # param: scopes
    # add any scopes you want to use with your API
    scopes=["openid", "product:buy"],
    binding_message=lambda product, quantity: f"Do you want to buy {quantity} {product}",
    user_id=lambda *_, **__: ensure_config()
    .get("configurable")
    .get("_credentials")
    .get("user")
    .get("sub"),
    # param: requested_expiry
    # Controls how long the authorization request is valid.
    # requested_expiry=301,

    # param: on_authorization_request
    # When this flag is set to `block`, the execution of the tool awaits
    # until the user approves or rejects the request.
    #
    # Given the asynchronous nature of the CIBA flow, this mode
    # is only useful during development.
    #
    # In practice, the process that is awaiting the user confirmation
    # could crash or timeout before the user approves the request.
    on_authorization_request="block",

    # requested_expiry=301,
)

This will intercept the tool call to initiate an Async Authorization request:

The Async Authorization request includes the user ID that will approve the request.
Auth0 sends the user a mobile push notification. The AI agent polls the /token endpoint for a user response.
The mobile application retrieves the bindingMessage containing the consent details, in this case, the details of the product to purchase.
The user responds to the request:
- If the request is approved, the tool execution will continue.
- If the request is rejected, the tool execution will not continue.

Pass credentials to the tools

Update the API route to pass the user session data to the agent in app/api/routes/chat.py:

app/api/routes/chat.py

# ...
from app.core.auth import auth_client
# ...

@agent_router.api_route(
    "/{full_path:path}", methods=["GET", "POST", "DELETE", "PATCH", "PUT", "OPTIONS"]
)
async def api_route(
    request: Request, full_path: str, auth_session=Depends(auth_client.require_session)
):
    try:
        # ... existing code

        # Prepare body
        body = await request.body()
        if request.method in ("POST", "PUT", "PATCH") and body:
            content = await request.json()
            content["config"] = {
                "configurable": {
                    "_credentials": {
                        "user": auth_session.get("user"),
                    }
                }
            }
            body = json.dumps(content).encode("utf-8")

            # ... existing code

Create a tool to call your API

In this example, we use a tool that buys products on the user’s behalf. When the user approves the transaction, the Auth0 AI SDK retrieves an access token to call the shop’s API. Upon completing the Async Authorization flow, the AI agent responds with a message confirming the purchase. The Auth0 AI SDK returns an error response if the user denies the transaction.Now, create a file app/agents/tools/shop_online.py and add the following code:

app/agents/tools/shop_online.py

import httpx
from langchain_core.tools import StructuredTool
from auth0_ai_langchain.async_authorization import get_async_authorization_credentials
from pydantic import BaseModel

from app.core.auth0_ai import with_async_authorization
from app.core.config import settings


class BuyOnlineSchema(BaseModel):
    product: str
    quantity: int


async def shop_online_fn(product: str, quantity: int):
    """Tool to buy products online."""

    api_url = settings.SHOP_API_URL

    if not api_url.strip():
        # No API set, mock a response
        return f"Ordered {quantity} {product}"

    credentials = get_async_authorization_credentials()

    if not credentials:
        raise ValueError("Async Authorization credentials not found")

    headers = {
        "Authorization": f"Bearer {credentials['access_token']}",
        "Content-Type": "application/json",
    }

    data = {
        "product": product,
        "quantity": quantity,
    }

    try:
        async with httpx.AsyncClient() as client:
            response = await client.post(
                api_url,
                headers=headers,
                json=data,
            )

        if response.status_code != 200:
            raise ValueError(f"Failed to buy product: {response.text}")

        return response.json()

    except httpx.HTTPError as e:
        return {
            "success": False,
            "error": f"Failed to buy product: {str(e)}",
        }


shop_online = with_async_authorization(
    StructuredTool(
        name="shop_online",
        description="Tool to buy products online.",
        args_schema=BuyOnlineSchema,
        coroutine=shop_online_fn,
    )
)

Update environment variables

You need to obtain an API Key from OpenAI to use an LLM.If you want to use an API, it must be registered with Auth0 and have a valid audience.Update the .env file with the following variables:

.env

# ... existing variables
# OpenAI API configuration
OPENAI_API_KEY="YOUR_API_KEY"

# Shop API configuration
SHOP_API_URL=<your-shop-api-url>
SHOP_API_AUDIENCE="https://api.shop-online-demo.com"

Update your app/core/config.py to include the shop audience:

app/core/config.py

# ...
class Settings(BaseSettings):
    # ... existing code
    SHOP_API_URL: str = ""
    SHOP_API_AUDIENCE: str = ""
    # ... existing code

Require async authorization for your tool

Call the tool from your AI app to make purchases. Update the app/agents/assistant0.py file with the following code:

app/agents/assistant0.py

# ...
from app.agents.tools.shop_online import shop_online

tools = [shop_online]

llm = ChatOpenAI(model="gpt-4.1-mini")

# ... existing code
agent = create_react_agent(
    llm,
    tools=ToolNode(tools, handle_tool_errors=False),
    prompt=get_prompt(),
)

Test your application

To test the application, start the FastAPI backend, LangGraph server, and the frontend:

In a new terminal, start the FastAPI backend:

cd backend
source .venv/bin/activate
fastapi dev app/main.py

In another terminal, start the LangGraph server:

cd backend
source .venv/bin/activate
uv pip install -U langgraph-api
langgraph dev --port 54367 --allow-blocking

This will open the LangGraph Studio in a new tab. You can close it as we won’t require it for testing the application.

In another terminal, start the frontend:

cd frontend
cp .env.example .env # Copy the `.env.example` file to `.env`.
npm install
npm run dev

Visit the URL http://localhost:5173 in your browser and interact with the AI agent.You can ask the AI agent to buy a product, for example, “Buy an XYZ phone.” Now, look for a push notification from the Auth0 Guardian app or your custom app integrated with the Auth0 Guardian SDK on your mobile device. Once you approve the notification, you should see the tool being executed and a response from the agent.That’s it! You’ve successfully integrated asynchronous authorization into your LangGraph FastAPI project.Explore the example app on GitHub.

Install dependencies

Install the following dependencies:

ai: Core Vercel AI SDK module that interacts with various AI model providers.
zod: TypeScript-first schema validation library.
dotenv: A module that loads environment variables from a .env file.
auth0: Auth0 Node.js library.
@ai-sdk/openai: OpenAI provider for the Vercel AI SDK.
@auth0/ai-vercel: Auth0 AI SDK for Vercel AI built for AI agents powered by the Vercel AI SDK.

Create a new Node.js project

npm init -y
npm i auth0@5 @auth0/ai-vercel@5 zod@3.25.76 ai@6 @ai-sdk/openai@3 dotenv@16

Add the below to package.json:

package.json

"main": "index.js",
"x-type": "module",
"scripts": {
  "start": "npx tsx src/index.ts"
},

Integrate Auth0 AI SDK

Integrate the Auth0 AI SDK into your application to secure your async AI agent workflow.

1. Create your environment file

In the root directory of your project, create the .env file and add the following variables. You can find your Auth0 domain, client ID, and client secret in the application you created in the Auth0 Dashboard.

.env

AUTH0_DOMAIN='<your-auth0-domain>'
AUTH0_CLIENT_ID='<your-auth0-application-client-id>'
AUTH0_CLIENT_SECRET='<your-auth0-application-client-secret>'

# You can use any provider of your choice supported by Vercel AI
OPENAI_API_KEY="YOUR_API_KEY"

# API
STOCK_API_URL=<your-stock-api-url>
STOCK_API_AUDIENCE=sample-stock-api

2. Require async authorization for your tool

To require asynchronous authorization for your tool wrap the tool with the Async authorizer, withAsyncAuthorization().Wrap the tool with the Async authorizer in src/lib/tools/buy.ts. This will intercept the tool call to initiate an Async Authorization request:

The Async Authorization request includes the user ID that will approve the request.
Auth0 sends the user a mobile push notification. The AI agent polls the /token endpoint for a user response.
The mobile application retrieves the binding_message containing the consent details, in this case, the quantity of stock to purchase for a stock ticker.
The user responds to the request:
- If the request is approved, the tool execution will continue.
- If the request is rejected, the tool execution will not continue.

In our example, we wrap a tool that buys shares for a given stock picker on the user’s behalf. When the user approves the transaction, the Auth0 AI SDK retrieves an access token to call the stock’s API. Upon completing the CIBA flow, the AI agent responds with a message confirming the purchase.The Auth0 AI SDK returns an error response if the user denies the transaction.Now, create and wrap the tool with the Async authorizer. Create a file src/lib/tools/buy.ts and add the following code:

src/lib/tools/buy.ts

import "dotenv/config";

import { tool } from "ai";
import { z } from "zod";

import { Auth0AI, getAsyncAuthorizationCredentials } from "@auth0/ai-vercel";
import { AccessDeniedInterrupt } from "@auth0/ai/interrupts";
export type Context = {
  userId: string;
};

const auth0AI = new Auth0AI();

export const buy = (context: Context) => {
  // highlight-next-line
  const withAsyncAuthorization = auth0AI.withAsyncAuthorization({
    // highlight-next-line
    userID: context.userId,
    // highlight-next-line
    bindingMessage: async ({ ticker, qty }) =>
      // highlight-next-line
      `Do you want to buy ${qty} shares of ${ticker}`,
    scopes: ["openid", "stock:trade"],
    audience: process.env["STOCK_API_AUDIENCE"]!,

    /**
     * Controls how long the authorization request is valid for.
     */
    // requestedExpiry: 301,

    /**
     * The behavior when the authorization request is made.
     *
     * - `block`: The tool execution is blocked until the user completes the authorization.
     * - `interrupt`: The tool execution is interrupted until the user completes the authorization.
     * - a callback: Same as "block" but give access to the auth request and executing logic.
     *
     * Defaults to `interrupt`.
     *
     * When this flag is set to `block`, the execution of the tool awaits
     * until the user approves or rejects the request.
     * Given the asynchronous nature of the CIBA flow, this mode
     * is only useful during development.
     *
     * In practice, the process that is awaiting the user confirmation
     * could crash or timeout before the user approves the request.
     */
    onAuthorizationRequest: async (_authReq, creds) => {
      console.log(
        `An authorization request was sent to your mobile device.`
      );
      await creds;
      console.log(`Thanks for approving the order.`);
    },

    onUnauthorized: async (e: Error) => {
      if (e instanceof AccessDeniedInterrupt) {
        return "The user has deny the request";
      }
      return e.message;
    },
  });

  return withAsyncAuthorization(
    tool({
      description: "Use this function to buy stock",
      inputSchema: z.object({
        ticker: z.string(),
        qty: z.number(),
      }),
      execute: async ({ ticker, qty }) => {
        const headers = {
          "Content-Type": "application/json",
        };
        const body = {
          ticker: ticker,
          qty: qty,
        };
        // highlight-next-line
        const credentials = getAsyncAuthorizationCredentials();
        const accessToken = credentials?.accessToken?.value;

        if (accessToken) {
          headers["Authorization"] = "Bearer " + accessToken;
        }

        console.log("Executing request to buy stock");

        const response = await fetch(process.env["STOCK_API_URL"]!, {
          method: "POST",
          headers: headers,
          body: JSON.stringify(body),
        });

        return response.statusText;
      },
    })
  );
};

3. Integrate the tool into an AI agent

Create a file at src/index.ts and add tool calling to the Vercel AI agent using the generateText() or streamText() functions:

src/index.ts

import { setAIContext } from "@auth0/ai-vercel";
import crypto from "node:crypto";
import { generateText, stepCountIs } from "ai";
import { openai } from "@ai-sdk/openai";

import { buy } from "./lib/tools/buy";

async function main() {
  const threadID = crypto.randomUUID();
  setAIContext({ threadID });

  const userPrompt = "Use the buy tool to purchase 3 shares of Google (GOOGL) stock";

  const { text } = await generateText({
    model: openai("gpt-4o-mini"),
    system: "You are an assistant. Invoke the tool if needed, then respond with a past-tense confirmation.",
    prompt: userPrompt,
    // pass an Auth0 user id. For example, 'auth0|100000000000000000000' or 'google-oauth2|100000000000000000000'
    tools: { buy: buy({ userId: "<authenticated-user-id>" }) },
    stopWhen: stepCountIs(3), // tool call + result + final assistant message
  });

  console.log("Final response:", (text && text.trim()) || "<empty>");
}

main().catch(console.error);

You should now have a CLI app that allows you to interact with the AI agent.

Test the application

To test the CLI app, set a prompt, such as “Buy 3 stocks of Google,” and pass it the user ID of the user approving or declining the transaction. You can get the user ID from the Auth0 Dashboard.To get the user ID, navigate to User Management > Users and click on a user. The ID should look like auth0|123456789.Now, run npm start and look for a push notification from the Auth0 Guardian app or your custom app integrated with the Auth0 Guardian SDK on your mobile device. Once you approve the notification, you should see the tool being executed on your console.Explore the example app on GitHub.

Prochaines étapes

Vous avez ajouté une étape d’autorisation pour protéger l’appel d’outils dans les agents d’IA asynchrones. Pour la suite :

Documentation sur l’autorisation asynchrone.
En savoir plus sur le flux Client-Initiated Backchannel Authentication.
Apprendre à configurer Rich Authorization Requests.
En savoir plus sur la trousse SDK Auth0 Guardian pour Android.
En savoir plus sur la trousse SDK Auth0 Guardian pour iOS.

​Choisissez votre pile technologique

​Download sample app

​Install dependencies

​Update the environment file

​Set up Human-in-the-Loop approvals

​Configure the Auth0 AI SDK

​Pass credentials to the tools

​Add Custom Authentication

​Create a tool to call your API

​Update environment variables

​Require async authorization for your tool

​Test the application

​Download sample app

​Install dependencies

​Update the environment file

​Set up Human-in-the-Loop approvals

​Configure the Auth0 AI SDK

​Require async authorization for your tool

​Update environment variables

​Integrate the tool into an AI agent

​Test the application

​Download sample app

​Install dependencies

​Update the environment file

​Set up Human-in-the-Loop approvals

​Configure the Auth0 AI SDK

​Pass credentials to the tools

​Create a tool to call your API

​Update environment variables

​Require async authorization for your tool

​Test your application

​Install dependencies

​Integrate Auth0 AI SDK

​1. Create your environment file

​2. Require async authorization for your tool

​3. Integrate the tool into an AI agent

​Test the application

​Prochaines étapes

Choisissez votre pile technologique

Download sample app

Install dependencies

Update the environment file

Set up Human-in-the-Loop approvals

Configure the Auth0 AI SDK

Pass credentials to the tools

Add Custom Authentication

Create a tool to call your API

Update environment variables

Require async authorization for your tool

Test the application

Download sample app

Install dependencies

Update the environment file

Set up Human-in-the-Loop approvals

Configure the Auth0 AI SDK

Require async authorization for your tool

Update environment variables

Integrate the tool into an AI agent

Test the application

Download sample app

Install dependencies

Update the environment file

Set up Human-in-the-Loop approvals

Configure the Auth0 AI SDK

Pass credentials to the tools

Create a tool to call your API

Update environment variables

Require async authorization for your tool

Test your application

Install dependencies

Integrate Auth0 AI SDK

1. Create your environment file

2. Require async authorization for your tool

3. Integrate the tool into an AI agent

Test the application

Prochaines étapes