Provisions the infrastructure needed to use the Gemini Enterprise Agent Platform (Vertex AI Agent Engine) on an existing GCP project: required APIs, service accounts, IAM bindings, a GCS artifacts bucket, and an Artifact Registry repository.

Prerequisites

Terraform >= 1.5.0
An existing GCP project
gcloud CLI authenticated with permissions to enable APIs and manage IAM

Usage

cp terraform.tfvars.example terraform.tfvars
# edit terraform.tfvars — set your project_id

terraform init
terraform plan
terraform apply

Variables

Name	Required	Default	Description
`project_id`	yes	—	Existing GCP project ID
`prefix`	no	`gemini`	Short prefix applied to all resource names

Outputs

Name	Description
`project_id`	GCP project ID
`agent_sa_email`	Runtime service account email (for apps)
`code_assist_sa_email`	Code Assist enterprise SA email

What gets created

7 GCP APIs enabled (aiplatform, cloudaicompanion, discoveryengine, dialogflow, secretmanager, iam, cloudresourcemanager)
2 service accounts — one for app runtime, one for IDE enterprise config
4 project IAM bindings

Setting up credentials

Use Application Default Credentials:

gcloud auth application-default login
gcloud config set project <your-project-id>

For workloads running on GCP (Cloud Run, GKE, Compute Engine), attach the service account to the resource — no credentials file needed.

Granting developer access to Gemini Code Assist

# Single user
gcloud projects add-iam-policy-binding PROJECT_ID \
  --member="user:YOU@DOMAIN" \
  --role="roles/cloudaicompanion.user"

# Google Group (recommended for teams)
gcloud projects add-iam-policy-binding PROJECT_ID \
  --member="group:devs@DOMAIN" \
  --role="roles/cloudaicompanion.user"

VS Code setup

Install the Gemini Code Assist extension from the VS Code Marketplace
Sign in with a Google account that has roles/cloudaicompanion.user on the project
In Settings, set Cloud AI Companion: Project to your project_id