GCP Computer
Sign InGet Started

Deployment

Deploy GCP Computer to Google Cloud Platform. The primary deployment path uses GCP Compute Engine with GitHub Actions CI/CD.

Deployment Options

  • GCP Compute Engine (primary) — full production path with VM-based sandboxing
  • Google Cloud Run — serverless container deployment (legacy)
  • Vercel — frontend-only deployment

Prerequisites

  1. A Google Cloud Platform project
  2. Enabled APIs:
    • Compute Engine API (compute.googleapis.com)
    • Cloud Run API (run.googleapis.com) — for Cloud Run path
    • Artifact Registry API (artifactregistry.googleapis.com)
  3. A service account with appropriate permissions

GCP Deployment

1. Provision Infrastructure

Run the provisioning script to set up the GCP project:

powershell -File scripts/gcp-provision.ps1

This script automates the creation of the service account, enables APIs, and sets up networking.

2. Configure Secrets

Add the following secrets to your GitHub repository (Settings > Secrets and variables > Actions):

| Secret | Description | |---|---| | GCP_SA_KEY | Full content of the service account JSON key | | GCP_PROJECT_ID | Your GCP project ID | | NEXTAUTH_SECRET | Random secret for session signing | | NEXTAUTH_URL | Deployed app URL | | GOOGLE_GENERATIVE_AI_API_KEY | Gemini API key |

3. Deploy

Push to the main branch. The GitHub Actions workflow automatically builds and deploys:

git push origin main

The deployment workflow:

  1. Checks out the repository
  2. Authenticates with GCP
  3. Builds the Docker image using the multi-stage Dockerfile
  4. Pushes the image to Artifact Registry
  5. Updates the Compute Engine VM or Cloud Run service

VM Setup (Compute Engine)

Use the VM setup script to configure a fresh Compute Engine instance:

bash scripts/vm-setup.sh

This script:

  • Installs Docker and dependencies
  • Clones the repository
  • Configures environment variables
  • Sets up Caddy as a reverse proxy
  • Starts the application

Persistent Database

Cloud Run is stateless. For production, use a managed PostgreSQL database:

Option A: Cloud SQL

Set up Google Cloud SQL for PostgreSQL and configure DATABASE_URL.

Option B: External Provider

Use Supabase, Neon, or Vercel Postgres.

Health Checks

After deployment, verify:

  • Login page loads
  • Authentication works
  • Sandboxes can be created
  • Commands execute successfully
  • Sessions persist across page reloads