Build a Google Chat app with an ADK AI agent

This page explains how to build a Google Workspace add-on that works in Google Chat and interfaces with an Agent Development Kit (ADK) AI agent hosted in Vertex AI Agent Engine .

Note: In Google Chat, add-ons appear to users as Google Chat apps. You can also build your Chat app using Google Chat API interaction events . To learn more, see the Extend Google Chat overview .

AI agents autonomously perceive their environment, reason, and execute complex, multi-step actions to achieve a defined goal. In this tutorial, you deploy the ADK LLM Auditor multi-agent sample that critiques and revises facts using Gemini and Google Search grounding.

The following diagram shows the architecture and messaging pattern:

In the preceding diagram, a user interacting with a Chat app implemented with an ADK AI agent has the following flow of information:

1. A user sends a message to a Chat app, either in a direct message or in a Chat space.
2. The Chat app logic that's implemented either in Apps Script or as a web server with HTTP endpoints receives and processes the message.
3. The AI agent that's implemented with ADK and hosted with Vertex AI Agent Engine receives and processes the interaction.
4. Optionally, the Chat app or AI agent can integrate with Google Workspace services, such as Calendar or Sheets, or other Google Services, such as Google Maps or YouTube.
5. The Chat app asynchronously sends responses using the Google Chat API to communicate the AI agent progress.
6. The responses are delivered to the user.

Objectives

• Set up your environment.
• Deploy the ADK AI agent.
• Deploy the Chat app.
• Configure the Chat app.
• Test the Chat app.

Prerequisites

• A Business or Enterprise Google Workspace account with access to Google Chat .
• A Google Cloud project with billing enabled. To check that an existing project has billing enabled, see Verify the billing status of your projects . To create a project and set up billing, see Create a Google Cloud project .

Set up your environment

Enable the Google Cloud APIs

Before using Google APIs, you need to turn them on in a Google Cloud project. You can turn on one or more APIs in a single Google Cloud project.

• In the Google Cloud console, enable the Google Chat, Vertex AI, and Cloud Resource Manager APIs.

  Enable the APIs

Configure the OAuth consent screen

All apps using OAuth 2.0 require a consent screen configuration. Configuring your app's OAuth consent screen defines what is displayed to users and app reviewers, and registers your app so you can publish it later.

1. In the Google Cloud console, go to Menu menu > Google Auth platform > Branding .

   Go to Branding

2. If you have already configured the Google Auth platform, you can configure the following OAuth Consent Screen settings in Branding , Audience , and Data Access . If you see a message that says Google Auth platform not configured yet , click Get Started :
1. Under App Information , in App name , enter a name for the app.
2. In User support email , choose a support email address where users can contact you if they have questions about their consent.
3. Click Next .
4. Under Audience , select Internal .
5. Click Next .
6. Under Contact Information , enter an Email address where you can be notified about any changes to your project.
7. Click Next .
8. Under Finish , review the Google API Services User Data Policy and if you agree, select I agree to the Google API Services: User Data Policy .
9. Click Continue .
10. Click Create .
3. For now, you can skip adding scopes. In the future, when you create an app for use outside of your Google Workspace organization, you must change the User type to External . Then add the authorization scopes that your app requires. To learn more, see the full Configure OAuth consent guide.

Create a service account in Google Cloud console

Create a new service account with the role Vertex AI User by following these steps:

Google Cloud console

1. In the Google Cloud console, go to Menu menu > IAM & Admin > Service Accounts .

   Go to Service Accounts

2. Click Create service account .
3. Fill in the service account details, then click Create and continue .
4. Optional: Assign roles to your service account to grant access to your Google Cloud project's resources. For more details, refer to Granting, changing, and revoking access to resources .
5. Click Continue .
6. Optional: Enter users or groups that can manage and perform actions with this service account. For more details, refer to Managing service account impersonation .
7. Click Done . Make a note of the email address for the service account.

gcloud CLI

1. Create the service account:

     gcloud iam service-accounts create  SERVICE_ACCOUNT_NAME 
    
   \ 
    --display-name="  SERVICE_ACCOUNT_NAME 
    
   " 
    
   

2. Optional: Assign roles to your service account to grant access to your Google Cloud project's resources. For more details, refer to Granting, changing, and revoking access to resources .

The service account appears on the service account page.

Create a private key

To create and download a private key for the service account, follow these steps:

1. In the Google Cloud console, go to Menu menu > IAM & Admin > Service Accounts .

   Go to Service Accounts

2. Select your service account.
3. Click Keys > Add key > Create new key .
4. Select JSON , then click Create .

   Your new public/private key pair is generated and downloaded to your machine as a new file. Save the downloaded JSON file as credentials.json in your working directory. This file is the only copy of this key. For information about how to store your key securely, see Managing service account keys .

5. Click Close .

For more information about service accounts, see service accounts in the Google Cloud IAM documentation .

Deploy the ADK AI agent

1. Open the LLM Auditor samplein the Vertex AI Agent Garden:

   Open the sample

2. Click Deploy.

3. If asked, select your Google Cloud project.

4. Click Deploy in Cloud Shell.

5. If asked, click Authorizeand go through the OAuth flow for the Cloud Shell.

6. When the Cloud Shell is fully loaded, press Enterto run the pre-loaded command line and start deploying.

7. If asked to enter region, press Enterto use default.

8. When complete, go to the Vertex AI Agent Engine:

   Open Vertex AI Agent Engine

9. Clickto copy the resource name of the newly deployed agent from the table.

Create and configure the Chat app project

1. Click the following button to open the ADK AI Agent QuickstartApps Script project.

   Open the project

2. Click info_outline Overview > Make a copy.

3. In your Apps Script project, click Project Settings > Edit script properties > Add script propertyto add the following script properties:

   1. REASONING_ENGINE_RESOURCE_NAME with the Vertex AI agent resource name copied in previous steps.
   2. SERVICE_ACCOUNT_KEY with the JSON key from the service account downloaded in previous steps such as { ... } .

4. Click Save script properties

5. In the Google Cloud console, go to Menu menu > IAM & Admin > Settings.

   Go to IAM & Admin Settings

6. In the Project numberfield, copy the value.

7. In your Apps Script project, click Project Settings.

8. Under Google Cloud Platform (GCP) Project, click Change project.

9. In GCP project number, paste the Google Cloud project number copied in previous steps.

10. Click Set project. The Cloud project and Apps Script project are now connected.

Create a test deployment

You need a deployment ID for this Apps Script project, so that you can use it in the next step.

To get the head deployment ID, do the following:

1. In the Chat app Apps Script project, click Deploy > Test deployments.
2. Under Head deployment ID, click Copy.
3. Click Done.

Configure the Chat app

Using your Apps Script deployment, follow these steps to deploy the Google Chat app for testing:

1. In the console , search for Google Chat API , and click Google Chat API.
2. Click Manage.

3. Click Configurationand set up the Chat app:

   1. In the App namefield, enter ADK Quickstart .
   2. In the Avatar URLfield, enter https://developers.google.com/workspace/add-ons/images/quickstart-app-avatar.png .
   3. In the Descriptionfield, enter ADK Quickstart .
   4. Under Functionality, select Join spaces and group conversations.
   5. Under Connection settings, select Apps Script project.
   6. In the Deployment IDfield, paste the Head deployment ID that you previously copied.
   7. Under Visibility, select Specific people and groups in your domain, and enter your email.

4. Click Save.

The Chat app is ready to respond to messages.

Test the Chat app

To test your Chat app, open a direct message space with the Chat app and send a message:

1. Open Google Chat using the Google Workspace account that you provided when you added yourself as a trusted tester.

   Go to Google Chat

2. Click add New chat .
3. In the Add 1 or more people field, type the name of your Chat app.

4. Select your Chat app from the results. A direct message opens.

5. In the new direct message with the app, type The Eiffel Tower was completed in 1900 and press enter .

   The Chat app replies with Critic and Reviser sub-agent responses.

To add trusted testers and learn more about testing interactive features, see Test interactive features for Google Chat apps .

Troubleshoot

When a Google Chat app or card returns an error, the Chat interface surfaces a message saying "Something went wrong." or "Unable to process your request." Sometimes the Chat UI doesn't display any error message, but the Chat app or card produces an unexpected result; for example, a card message might not appear.

Although an error message might not display in the Chat UI, descriptive error messages and log data are available to help you fix errors when error logging for Chat apps is turned on. For help viewing, debugging, and fixing errors, see Troubleshoot and fix Google Chat errors .

Clean up

To avoid incurring charges to your Google Cloud account for the resources used in this tutorial, we recommend that you delete the Cloud project.

1. In the Google Cloud console, go to the Manage resources page. Click Menu menu > IAM & Admin > Manage Resources .

   Go to Resource Manager

2. In the project list, select the project you want to delete and then click Delete delete .
3. In the dialog, type the project ID and then click Shut down to delete the project.

Related topics

• Fact-check statements with an ADK AI agent and Gemini model
• Plan travels with an AI agent accessible across Google Workspace
• Integrate fundamental AI concepts in Chat apps
• Answer questions based on Chat conversations with a Gemini AI Chat app
• Respond to incidents with Google Chat, Vertex AI, Apps Script, and user authentication
• Manage projects with Google Chat, Vertex AI, and Firestore