Google Cloud Run vs while: which one to choose when your AI agent deploys?
Cloud Run runs your containers on Google’s infrastructure, to be assembled with Cloud SQL, IAM and the rest. With while, your AI agent deploys the app, its database and its access in a single request, over MCP, and sensitive actions, like deleting a production database, wait for your approval.
By Grégoire Mougeolle, co-founder of Nextmap. Facts checked on November 2, 2026. How this comparison is made.
What changes between Google Cloud Run and while
| What matters | Google Cloud Run | while |
|---|---|---|
| Who it’s for | Teams that assemble their app in Google Cloud: containers, Cloud SQL, IAM1 | People who build their apps with an AI agent, solo or as a team |
| Your AI agent | An MCP server that deploys, with no delete tool2 | Connected over MCP, with OAuth: it deploys, creates the database, reads the logs and fixes things |
| What protects production | The Google Cloud permissions of the signed-in person2 | Sensitive actions, like deleting a production database, wait for a person |
| The database | Cloud SQL, set up separately, with point-in-time recovery turned off by default when the database is created from the command line or the API, as an agent would do it3 | PostgreSQL in while, backed up every day, with point-in-time recovery |
| Who opens your app | A Google account or the company’s SSO, through Identity-Aware Proxy4 | Everyone, a password, or the company account with the Pro plan |
| In France | Yes, a region in France5 | Yes, with Nextmap, a French company |
| To get started | A free tier every month, then billing by the second, with the Cloud SQL database on top6 | A trial without an account. Then €0 to get started, and €13.75 per month for an app and its database |
Try while on your app, without an account. Copy the instructions
Cloud Run, Google’s power with no server to manage
Cloud Run deploys your code with a single command, in more than 20 regions at once, and bills only for the time it runs1. It can also serve AI models on GPUs, which start in 5 seconds and stop when no one is using them1.
With while, your AI agent does the rest
- No server to maintain. There’s no system to update, no disk to watch and no certificate to renew: your agent deploys, and the rest is taken care of.
- Your databases are backed up for you. Every day, with point-in-time recovery, and a production backup is tested every day.
- Nothing serious happens without you. Deleting a production database waits for your approval, before it runs, and every action is recorded in a signed activity log.
- Your keys stay secret. Your app uses them, your agent never sees them.
- Your colleagues sign in with the company account. Google Workspace or Microsoft 365, on the Pro plan, with no sign-in page to code.
Moving from Google Cloud Run to while
Take over your app
You, to your agent: Move our Cloud Run service to while.
Your agent takes over your code or your image, writes while.yaml and deploys the app.
Copy your Cloud SQL database
You, to your agent: Copy our Cloud SQL database to while.
Your agent exports the database with pg_dump, then loads it into while.
Carry over your variables
You, to your agent: Carry over our environment variables.
Your agent carries your settings over to while. You enter your secret keys yourself: it never sees them.
Which one to choose
- Choose Google Cloud Run if your team already lives in Google Cloud, or if you need GPUs to serve a model.
- Choose while if your AI agent is the one deploying, and you want the app, the database and access in one go, in France, without breaking anything.
Next steps
Sources
- Google Cloud Run, overview, accessed November 2, 2026
- Cloud Run, the remote MCP server, accessed November 2, 2026
- Cloud SQL for PostgreSQL, backups and point-in-time recovery, accessed November 2, 2026
- Cloud Run, Identity-Aware Proxy, accessed November 2, 2026
- Cloud Run, regions, accessed November 2, 2026
- Cloud Run, pricing, accessed November 2, 2026