## The short version

Ample, Railway, Netlify, Bolt, Replit and Google Cloud Run can stop spending at a limit; Vercel can pause production if you turn that on; Render, Fly.io, Heroku, DigitalOcean, Cloudflare, Northflank and Koyeb document only alerts or nothing. Ample's Free plan has the cap on by default, so an unattended agent cannot run up a bill.

A coding agent that deploys on its own can also create resources on its own. A spending cap is what turns a runaway loop into a paused app instead of an invoice. Most platforms offer alerts; fewer actually stop.

## Who stops, who alerts

| Platform | What happens at the limit | On by default? |
| --- | --- | --- |
| [Ample](/content/pricing/index.html) | Apps stop serving until the account is back under budget; nothing is deleted and nothing is billed | Yes, on Free ($5 a month, $2 a day) |
| [Railway](https://docs.railway.com/pricing/cost-control.md) | A hard usage limit (minimum $10) takes every workload offline; services come back when it is raised | No, you set it |
| [Netlify](https://docs.netlify.com/manage/accounts-and-billing/billing/resume-paused-projects/) | When credits run out, every project on the team pauses until the next cycle | Yes, unless auto recharge is on |
| [Replit](https://docs.replit.com/billing/managing-spend) | A usage limit blocks usage-based services; a service shutdown limit suspends them | No, you set it |
| [Bolt](https://support.bolt.new/bolt-cloud/hosting/plans.md) | Free sites stop at the allowance; on Pro the site pauses at a spending cap you set | Yes on Free; you set it on Pro |
| [Google Cloud Run](https://docs.cloud.google.com/billing/docs/how-to/budgets-spend-caps) | Spend cap budgets (generally available since October 13, 2026) make services return errors until the cap lifts; reporting delay means you can still go over | No, you set it |
| [Vercel](https://vercel.com/docs/spend-management) | Pro sends alerts; an optional switch pauses production for every project | Alerts only, unless enabled |
| [Render](https://render.com/docs/build-pipeline.md) | Only build minutes have a hard limit; services keep running and are billed | No compute cap documented |
| [Fly.io](https://docs.fly.io/about/cost-management) | "If you go over, we'll bill you"; no billing alerts yet | No |
| [Heroku](https://devcenter.heroku.com/articles/limits) | No spending cap or alert found; billed in arrears | No |
| [DigitalOcean](https://docs.digitalocean.com/platform/billing/spend-alerts/) | Spend alerts only: "A spend alert budget is not a spending cap" | No |
| [Cloudflare](https://developers.cloudflare.com/billing/manage/budget-alerts/) | Budget alerts are informational and do not pause usage; per-Worker CPU limits are the guard | No |
| [Northflank](https://northflank.com/docs/v1/application/billing/pricing-on-northflank) | Spend alerts; no hard cap documented | No |
| [Koyeb](https://www.koyeb.com/docs/faqs/pricing) | No spending limits yet | No |

## What to look for

- **A cap that stops, not just warns.** Alerts at 50%, 75% and 100% are useful, but an agent running at 3 a.m. will not read them.
- **What stops.** Netlify and Railway take everything offline at once; Vercel's switch pauses production deployments; Ample stops serving the account's apps and keeps their data.
- **How it comes back.** Railway restarts services when you raise the limit; Vercel needs each project resumed by hand; Ample serves again on the next request once the account is under budget.

## Sources

Each row links to that platform's own documentation, checked on October 14, 2026. Ample's behaviour is described on the [pricing page](/content/pricing/index.html).