Using Google Cloud Credits in VS Code

What you will end up with

Gemini models listed in the Copilot Chat model picker in VS Code, billed to your own Google Cloud project and paid for by your education credits. You will set things up once in the Google Cloud Console, download a key file, and hand that key to a VS Code extension. You do not need to install the Google Cloud command-line tool (gcloud).

You will need:

Your reed.edu Google account.
The coupon code for the education credits.
VS Code (version 1.110 or newer) with Copilot Chat, signed in with your GitHub account.

Part 1: Google Cloud

Step 1: Redeem the education credits

Open the credit application link you were given. Fill in your name and make sure the account email is your reed.edu address. Enter the coupon code and accept the terms.

The GCP credit application form, signed in with a reed.edu account, with a field for the coupon code.

Watch out for

Use your reed.edu account, not a personal Gmail. Otherwise the credits will not be connected to the Reed organization you pick in Step 2.
The coupon code is single-use. Do not share it.

Step 2: Create a project

Create a new project in the Google Cloud Console. Give it a name (for example "ILCP 2026"), and choose reed.edu as the organization and as the location.

New Project creation form — The New Project form: project name ILCP 2026, organization reed.edu, location reed.edu.

Google turns the name into a Project ID (here ilcp-2026). Write the Project ID down. You will type it into VS Code in Step 8.

Watch out for

The Project ID cannot be changed later. If you want a different one, click Edit next to it before clicking Create.

Step 3: Enable the Agent Platform API

Go to APIs & Services > Library, search for "Agent Platform API", open it, and click Enable. Make sure your new project is selected at the top of the page first.

API Library search results showing the Vertex AI to Agent Platform rename banner — The API Library. A banner notes that Vertex AI is now called Agent Platform.

Agent Platform API details page showing status Enabled — The Agent Platform API page, showing Status: Enabled.

Watch out for

Google renamed Vertex AI to "Gemini Enterprise Agent Platform" in 2026. Older tutorials, and some menus, still say Vertex AI. They are the same thing.
This is the only API you need. If the console later suggests enabling a long list of other APIs, you can ignore it.

Step 4: Create a service account

A service account is a separate identity that VS Code uses to call the models for you. Go to IAM & Admin > Service Accounts and click Create service account. Name it something like vscode. On the Permissions step, give it the role Agent Platform User. You can skip the last optional step.

Service account details page for vscode2, showing its name, email and Enabled status — The Service account details page, under IAM & Admin > Service Accounts.

Create service account, Permissions step, role selector open — The Permissions step of Create Service Account, where you choose the role.

Create service account, Permissions step, with Agent Platform User selected as the role — With **Agent Platform User** chosen as the role. Click **Done**.

Create service account, final step, confirmation the account and policy were created — After clicking Done: "Service account created" and "Policy updated. It may take a few minutes for these changes to become active."

Watch out for

The role must be Agent Platform User (its technical name is roles/aiplatform.user). Without it, no models will appear.
As the message says, the role can take a few minutes to start working.

Step 5: Download a key for the service account

Click the new service account, open the Keys tab, and choose Add key > Create new key > JSON. A .json file downloads to your computer.

Service accounts list for project ILCP 2026, with the new service account circled — The Service accounts list. Click the email of the account you just created.

Private key saved to your computer dialog — "Private key saved to your computer," with the name of the downloaded JSON file.

Watch out for

This file works like a password that spends your credits. Never email it, share it, or put it in a Git repository or a published web folder.
You will import it into VS Code in Step 7 and can delete it after that.

Part 2: VS Code

Step 6: Install the extension

In VS Code, open the Extensions panel and search for "google agent". Install Google Agent Platform for Copilot Chat (Vertex AI Models Chat Provider) by jorsm.

VS Code Extensions marketplace search for google agent — Searching the Extensions panel for "google agent". The one to install is "Google Agent Platform for Copilot Chat (Vertex AI Models Chat Provider)" by jorsm.

Watch out for

The same author has an older extension called "Vertex AI Models Chat Provider (Deprecated)". Do not install that one.
Ignore any instructions that appear in the chat panel about API keys or pip install. That is a different method.

Step 7: Import the key

Open the Command Palette: Cmd+Shift+P (Mac) or Ctrl+Shift+P (Windows).
Type Google Agent Platform: Import Service Account JSON File and press Enter.
Choose the JSON file you downloaded in Step 5.
When asked for a name, keep the suggested one (your Project ID) and press Enter.

VS Code Command Palette filtered by google, with Google Agent Platform: Import Service Account JSON File highlighted — Typing `google` in the Command Palette lists the extension's commands. Pick **Import Service Account JSON File**.

You should see: "Service Account '...' securely stored and activated for this workspace." The key is now saved inside VS Code, so you can delete the JSON file from your Downloads folder.

Watch out for

The key is turned on only for the folder (workspace) you have open. When you open a different folder, run Google Agent Platform: Select Authentication Method once and pick your stored key from the list. You do not need to import the file again.

Step 8: Enter your Project ID

Open Settings, in one of these ways:
- If a notification says "'vertexAiChat.projectId' is required to call Vertex AI", click its Open Settings button.
- Use the keyboard: hold Cmd and press the comma key (Mac), or hold Ctrl and press the comma key (Windows).
- Use the menu: Code > Settings > Settings (Mac) or File > Preferences > Settings (Windows).
Make sure the User tab is selected, so the setting applies to every folder.
Search for vertexAiChat.projectId.
Type your Project ID from Step 2 (for example ilcp-2026).

VS Code notification: Google Agent Platform: 'vertexAiChat.projectId' is required to call Vertex AI, with an Open Settings button — If the Project ID is missing, this notification appears. **Open Settings** takes you to the right place.

Watch out for

Use the Project ID, not the project name. "ILCP 2026" will not work; "ilcp-2026" will.

Step 9: Pick a model in Copilot Chat

Open the Chat panel. Check the indicator at the bottom left of the panel: it must say Local. A Local session shows the modes Agent / Ask / Plan with the model name next to them.

Chat panel in a Local session in Agent mode, with the model picker open on the Google Agent Platform tab listing Gemini models — The correct setup: Local at the bottom left, the mode (Agent), and the model name next to it. Here the model picker is open on the Google Agent Platform tab, listing Gemini models.

Click the model name (it says Auto at first). At the top of the list, click the star icon next to Copilot to see the models from your Google Cloud project. Choose the newest Gemini Flash model (for example, Gemini 3.8 Flash) and send a test message.

Watch out for

Stay with a Flash model. The Pro models (for example, Gemini 3.1 Pro) cost much more per message and can use up your credits quickly.
If the list is empty, run Google Agent Platform: Refresh Models from the Command Palette (Cmd+Shift+P on Mac, Ctrl+Shift+P on Windows). Finding the models can take up to a minute.

VS Code Command Palette filtered by refresh, listing Google Agent Platform: Refresh Models among other commands — Type `refresh` in the Command Palette and pick **Google Agent Platform: Refresh Models**. The highlighted first item is a different command.

Making the credits last

Use a Gemini Flash model by default. Flash models cost much less per message than Pro models or Claude. Switch to a bigger model only when you need it.
Check your spending. Run Google Agent Platform: Show Usage & Costs Dashboard to see an estimate. Your Google Cloud Billing page shows the real numbers.
Start new chats often. Long conversations resend everything each time, which costs more.

If something goes wrong

A notification says credentials expired, with a "Login with gcloud" button

This means VS Code is not using your key in the folder you have open. Do not click "Login with gcloud" (it needs gcloud, which you did not install). Instead, run Google Agent Platform: Select Authentication Method and pick your stored key.

VS Code notification: Google Cloud credentials have expired or are invalid, with a
Login with gcloud button — The notification you see when no key is selected for the current folder.

No Google models after clicking the star icon

Check the Project ID in Settings (Step 8).
Check that the key is selected for this folder (Step 7).
Check that the service account has the Agent Platform User role (Step 4).
If you just gave the role, wait about 10 minutes, then run Google Agent Platform: Refresh Models.

The chat shows Interactive / Plan / Autopilot instead of Agent / Ask / Plan

The session is not Local, so your Google models cannot appear. Open the session menu above the chat input box, choose Continue In > Local, and check that the indicator at the bottom left says Local.

Chat panel not in Local mode, showing the Interactive / Plan / Autopilot selector — Wrong setup: the session is set to Copilot, so the modes read Interactive / Plan / Autopilot and your Google models are not available.

Gemini works but Claude models do not

Claude models must be enabled separately. In the Google Cloud Console, open Model Garden, find the Claude model you want, and click Enable. Gemini models do not need this step. Claude models also cost more, so use them sparingly.

Checklist

Redeem the credits with your reed.edu account.
Create a project and write down its Project ID.
Enable the Agent Platform API.
Create a service account with the Agent Platform User role.
Download a JSON key for it.
Install "Google Agent Platform for Copilot Chat" by jorsm.
Import the JSON key, then delete the file.
Set vertexAiChat.projectId in User settings.
In a Local chat session, click the model name, then the star icon, and pick the newest Gemini Flash model.