Replit deployment costs: Autoscale vs Reserved VM vs Static, and how to cap your spend

Publishing on Replit is one click, but the bill depends on the deployment type, your database and how much Agent work shares the same credits. Here's what each piece costs and how to put a hard limit on it.

Short answer

Agent-built Replit apps run on Autoscale (a monthly base fee plus compute units and requests, nothing while idle) or a Reserved VM (a fixed monthly price); Static is nearly free but doesn't work for Agent apps. The production database bills separately for compute hours and storage. Everything comes out of your plan's monthly credits first, then is billed as usage, so set a usage limit in Settings → Usage and cap Autoscale's max machines before you share the link.

What each deployment type costs

Replit bills publishing per deployment type. All costs come out of your monthly credits first; you pay usage-based fees only after those credits are used up, and Replit says unused credits don't roll over to the next month (Pro credits roll over for two months).

Replit publishing costs by deployment type
Type Price You pay Best for
Autoscale $2/month base fee + $0.60 per million compute units + $0.40 per million requests Only while requests are being served; scales to zero when idle Web apps and APIs, including Agent-built apps (the default)
Reserved VM Shared 0.5 vCPU / 2 GB: $15/month. Dedicated 1 vCPU / 4 GB: $35, 2 vCPU / 8 GB: $50, 4 vCPU / 16 GB: $130 per month Fixed monthly price; the server never sleeps Bots, background workers, always-on APIs
Scheduled $2/month base fee + $0.60 per million compute units; the scheduler is free For the duration of each run Backups, reports, notifications
Static Hosting free; $0.05 per GB of data transfer Only for the data your site serves Landing pages and docs; not compatible with Agent-built apps
Production database (PostgreSQL) $0.16 per compute hour + $0.35 per GiB-month of storage; development databases are free While the database receives requests, plus 5 minutes after the last one Any published app that stores data in Replit's database
  • Compute units: 1 CPU-second = 18 units, 1 GB-second of RAM = 2 units.
  • Core includes 100 GiB of outbound data transfer per month; Pro's allowance is on the pricing page.
  • Replit's Aug 2026 price-update table lists the Autoscale base fee as $1/month, while the Publishing costs page (used here, and in Replit's own examples) shows $2/month.

Prices in USD before tax, checked 17 Oct 2026 on Replit: Publishing costs (opens in a new tab).

How Autoscale billing works

  • The app starts when the first request arrives, serves it, and goes idle after 15 minutes without traffic. You pay for CPU and memory only while requests are being processed; Replit says billing time is often 1 to 2 seconds per request.
  • Requests that arrive at the same time share the same compute, so you pay for the total time the server is working, not once per request.
  • Max machines (Publishing → Adjust settings → Machine configuration) is the upper limit of servers Autoscale can add. Replit says your app never scales beyond it, which caps your maximum cost. The pane shows the hourly cost per machine and at max traffic.
  • Replit's own estimates, before credits: a small business site with 75,000 requests a month at about $2.39, an API with 300,000 requests at about $4.50, and a daily Scheduled job at about $2.03 per month.

Which deployment type fits a vibe-coded app?

  • Built with Replit Agent: Autoscale (the default) or Reserved VM. Replit says Static deployments are not compatible with Agent-built apps because they need a backend server.
  • Low or bursty traffic: Autoscale, because idle time costs nothing to run.
  • A bot, websocket server or worker that must stay connected: Reserved VM, a fixed monthly price for a server that never sleeps.
  • A nightly job (backups, emails, reports): Scheduled. It runs your command on a cron schedule and has no public URL.
  • A plain HTML, CSS and JavaScript site you wrote yourself: Static, which only charges for data transfer.

The database is billed separately

  • Development databases are free. Production databases bill for compute time (while they receive requests, plus 5 minutes after the last one) and for storage, measured as the month's maximum. Each PostgreSQL database takes 33 MB even when empty.
  • Production database storage is limited to 100 GiB per database on paid plans and 1 GiB on Starter.
  • An app that's queried around the clock keeps its database active around the clock, so check the database line on the Usage page, not just the deployment.

How to cap your spend

  • Set a usage limit: Settings → Usage → Account usage → Manage limits (organizations: Usage → Manage limits). When you reach it, usage-based services are blocked until the next billing cycle or until you raise it.
  • In the same dialog, a service shutdown limit suspends services when it's reached, so they can't keep adding charges.
  • Lower Autoscale's max machines to put a ceiling on what a traffic spike can cost.
  • Credit packs (Core and Pro) expire six months after purchase. Auto-reload buys a new pack when your balance runs low; turn it off in Settings → Billing if you want spending to stop at the credits you have. Replit says limits apply after pack credits are used, and that setting limits to $0.01 restricts you to the credits you bought.
  • Watch the Usage page: it updates in near real time and filters by resource, project and date.
  • On Core and Pro, Agent work uses the same monthly credits as publishing, so a heavy building week leaves less to cover hosting.

Replit's docs explain how to set these limits but don't say any limit is set for you, and Reddit threads describe Autoscale bills that grew over a weekend. Set the limit before you share the link.

What the free Starter plan covers

  • One published app, which goes down after 30 days (you can publish it again) and shows a “Made with Replit” badge.
  • Monthly credits for production databases, object storage and publishing (Autoscale and Static).
  • More published apps, badge removal and connectors such as Stripe need Core. See our custom domain guide for what that means for your own domain.

Moving off Replit to cut costs

If a Reserved VM is what you need, compare it with an always-on host before you commit. Replit Secrets and Replit's database don't travel with the code, so recreate environment variables and migrate data first. Our Replit → Railway and Replit → Render guides cover the move.

Common questions

Is Replit hosting free?

The Starter plan includes one published app, but it goes down after 30 days and carries a badge. Static deployments have free hosting and charge only for data transfer, but they don't work for apps built with Replit Agent. Anything lasting runs on Core or Pro, where publishing costs come out of your monthly credits first.

Why did my Replit bill go over my plan price?

Publishing, the production database, data transfer and Agent work all draw on the same monthly credits. Once they're used up, Replit bills the rest as usage. A usage limit in Settings → Usage stops that.

Does Autoscale cost money when nobody visits?

Replit says Autoscale scales to zero after 15 minutes without requests and you don't pay for compute while it's idle. The monthly base fee still applies, and a production database bills separately whenever it's active.

Is a Reserved VM cheaper than Autoscale?

Only if your app is busy most of the time or must stay connected. A Reserved VM is a fixed monthly price whether or not anyone visits; Autoscale charges only while requests are being served.

How do I stop Replit charging me at all?

Set your usage limit (or service shutdown limit) to the lowest value, turn off credit pack auto-reload, and unpublish deployments you don't need. Replit notes that continuous payment failures can lead to published apps being suspended.

Related on VibeLiftoff