# Replit

Official pricing: https://docs.replit.com/billing/deployment-pricing  
Category: paas · Isolation: vm

## Pricing regimes (raw)

- **Reserved VM deployment (always-on shared/dedicated VM, hourly)** (sizes): shared 0.5 vCPU / 2 GiB 0.5 vCPU/2 GiB $0.0208/h; dedicated 1 vCPU / 4 GiB 1 vCPU/4 GiB $0.0486/h; dedicated 2 vCPU / 8 GiB 2 vCPU/8 GiB $0.0694/h; dedicated 4 vCPU / 16 GiB 4 vCPU/16 GiB $0.1806/h; dedicated 8 vCPU / 32 GiB 8 vCPU/32 GiB $0.3611/h; dedicated 16 vCPU / 64 GiB 16 vCPU/64 GiB $0.7222/h
- **Autoscale / Scheduled deployment (compute units while serving or running)** (resource): $0.03888/vCPU-h, $0.00432/GiB-h [alt]

## Features

Yes: snapshots, fork/clone, persistent disk, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, full VM (own kernel), desktop GUI, preinstalled agents, SSH, HTTPS preview URLs, custom domains, open internet, static egress IP, SOC 2, SSO, Python, Node.js, Go, Java, wake on request, scheduled runs (cron), audit logs, EU data residency, spend limits, MCP server, automatic snapshots, agent harness API, gVisor or VM (no shared kernel), inbound access rules, secret proxy

No: memory snapshots, pause/resume, volumes, your own Docker/OCI image, Docker inside, nested virtualization, systemd, browser, computer-use API, code interpreter, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, public IPv4, egress allowlist, self-hosting, BYOC, open source, HIPAA, GPU, Windows, macOS, GPU desktop, HTTP method/path egress rules, live resize, live fork (no pause), memory fork, snapshots on demand, hosted agent API (their own agent), secret proxy for any API, extra volumes, shared volumes

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/replit.json → feature_evidence

## Caveats

- Replit is an AI app builder / IDE + hosting platform, not a programmatic sandbox API. Reserved VM = one always-on published app; there is no API to spin up N ephemeral VMs.
- Development workspace time is 'Unlimited' on Core/Pro and not billed per hour, but workspace vCPU/RAM per plan is no longer published (usage docs say 'determined by plan'). THIRD-PARTY ONLY: Core workspace 4 vCPU / 8 GiB, Starter 2 GiB RAM (nocode.mba / eesel.ai / p0stman.com). Hard limit: 20 concurrent Replit Apps per account. Not modelled as a mode.
- Agent usage is effort-based (billed per checkpoint by task complexity, at provider API rates) and is deducted from the same credits; Core/Pro get a daily 'Free Mode' allowance. Not modelled.
- Reserved VM hourly rates come from the Aug-2026 update table; the deployment-pricing page lists monthly prices ($15/$35/$50/$130) that equal ~720 h. Engine at 730 h gives ~1.4% more. Whether a Reserved VM can be stopped to avoid charges is undocumented, so reserved-vm has requires_always_on=true (verifier 2026-09-28; before, the engine billed it per session hour, the 'unverified' $1,589 scenario in the regime file, instead of the documented always-on $6,500).
- Autoscale base fee conflict (re-confirmed 2026-09-28): deployment-pricing page constants AutoscaleBaseFee '$2' / AutoscaleDeployment '$2.00' vs the Aug-2026 update table 'Autoscale base fee $1 / mo' in both the current and new columns. Unresolved; base fee not modelled (start_fee 0). Autoscale machine sizes and max machines not published. Request fee $0.40/1M not modelled.
- Pro tiers verified from the rendered pricing page 2026-09-29: $100 ($90/mo annual), $250 ($215), $500 ($425), $1,000 ($825), $2,500 ($2,000). The selector labels each tier by its credit amount ('$100 credits'), and the FAQ says 'Pro subscribers get at least $100 in monthly credits'.
- Aug 2026 price cut (effective first billing cycle on/after 2026-08-01): CU $3.20 -> $0.60/1M, requests $1.20 -> $0.40/1M, outbound $0.10 -> $0.05/GiB, App Storage $0.03 -> $0.015/GiB-mo, DB storage $1.50 -> $0.35/GiB-mo; Reserved VMs changed per size (dedicated 4/16 $0.22 -> $0.1806/h; legacy shared 1/4 rose $0.042 -> $0.0486/h).
- Outbound: 100 GiB/month included for deployments on Core and Pro (plus 100 GiB Core / 1,000 GiB Pro for development), then $0.05/GiB; ingress free. Egress free allowance encoded per account.
- Storage: no snapshot/paused-state billing (snapshot_gib_month null). App Storage (object) $0.015/GiB-mo; production PostgreSQL $0.16/compute-hour (suspends after 5 min idle) + $0.35/GiB-mo, 10 GiB limit per DB.
- Credit packs $100-$1,000 (up to $50 off), expire after 6 months; budgets are hard caps that suspend usage-based services.
- Pro tier credits assumed equal to the monthly price of each tier (pricing page shows '$100 credits' on the tier selector; higher tiers' credit amounts not explicitly restated).
- Regions: North America, Europe, Asia, South America, Australia selectable on Core/Pro/Enterprise (permanent per project); no regional price differences published.
- No benchmark data (perf null). Isolation 'container' is unverified; published apps run on GCP.
- Unverified (2026-09-28): Autoscale compute units are described both as 'CPU time: Processing power used' and as 'based on the CPU and memory you choose'; card keeps alloc (configured size x request-serving time). If Replit meters actual CPU seconds, a mostly-idle-while-serving app costs less on CPU.
- SSH (verified 2026-09-28, docs.replit.com/replit-workspace/ssh): SSH connects to a Replit App's development workspace and is 'available for Core, Pro, and all other paid plans'; SSH into published deployments (Reserved VM / Autoscale) is not documented, so those regimes carry ssh = null. Card-level ssh = true refers to the (unmodelled) workspace.
- App Storage objects have a 7-day minimum billing period; App Storage data transfer is $0.05/GiB (source: docs.replit.com/billing/object-storage-billing).
- Credit packs (Core/Pro): $100 for $100, $300 for $290, $500 for $480, $1,000 for $950; they expire 6 months after purchase, used earliest-expiring first, with optional auto-reload.

## How this provider charges


Replit is an AI app builder + cloud IDE + hosting platform. Everything is paid from one pool of **credits**: the plan fee
converts into monthly credits (Core $20, Pro $100-$2,500), which pay for Agent work, deployments, databases, storage
and transfer; beyond that, pay-as-you-go up to a budget. Development workspace time is "Unlimited" and not metered.
Compute you can size and price is **deployments**: Reserved VM (always-on, hourly), Autoscale and Scheduled (compute
units), Static (transfer only). A large price cut took effect with the first billing cycle on/after 2026-08-01.

Verified live 2026-09-28: replit.com/pricing (HTML), docs.replit.com billing pages listed below.

## Regime table

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| Starter (free) | Default | $0; daily Agent credits (monthly cap) + small monthly cloud credits (amounts unpublished) | 1 published app, expires after 30 days; North America only; no full build mode | https://docs.replit.com/billing/plans/starter-plan, https://docs.replit.com/features/publishing/project-geography.md |
| Core | Personal builders | $20/mo ($18 annual); fee = $20 monthly credits, reset monthly | 5 collaborators; 100 GiB dev + 100 GiB deploy outbound; 50 GiB storage/app; Free Mode chat up to 30 h / 60 projects | https://replit.com/pricing |
| Pro tiers | Teams up to 15 builders | Monthly $100 / $250 / $500 / $1,000 / $2,500 (annual $90 / $215 / $425 / $825 / $2,000); credits = tier; **roll over 2 months** | 10 parallel agents; 1,000 GiB dev outbound; 100 GiB deploy outbound; 256 GiB storage/app; 28-day DB rollback; 50 viewers | https://replit.com/pricing, https://docs.replit.com/billing/plans/replit-pro |
| Enterprise | SSO, single tenant, static outbound IPs | Custom (sales) | custom limits | https://replit.com/pricing |
| Pay-as-you-go overage | Credits exhausted | Same unit rates; stops at budget (hard cap suspends usage-based services) | — | https://replit.com/pricing (FAQ), https://docs.replit.com/billing/managing-spend.md |
| Credit packs | Prepay | $100-$1,000 packs, up to $50 off; expire 6 months; optional auto-reload | — | https://docs.replit.com/billing/managing-spend.md |
| Reserved VM | Always-on app (bots, APIs, workers) | Hourly while deployed (~720 h = listed month) | shared 0.5/2 GiB $0.0208/h ($15/mo); dedicated 1/4 $0.0486 ($35); 2/8 $0.0694 ($50); 4/16 $0.1806 ($130); 8/32 $0.3611; 16/64 $0.7222 | https://docs.replit.com/billing/aug-cloud-billing-updates.md, https://docs.replit.com/billing/deployment-pricing |
| Autoscale | Variable-traffic web apps | Base fee/deployment + CU while serving requests + per request; scale to zero, idle after 15 min | $0.60/1M CU; 1 CPU-s = 18 CU ($0.03888/vCPU-h), 1 GB-s = 2 CU ($0.00432/GiB-h); $0.40/1M requests; base **$2/mo** (pricing page) vs **$1/mo** (Aug table) — conflict | same |
| Scheduled | Cron/background jobs | Base fee + CU during runs; scheduler free | $2/mo + $0.60/1M CU | https://docs.replit.com/billing/deployment-pricing |
| Static | Static sites | Hosting free; transfer only | $0.05/GB | same |
| Outbound transfer | Beyond included allowance | Per GiB egress; ingress free | $0.05/GiB (was $0.10 before Aug 2026) | https://docs.replit.com/billing/aug-cloud-billing-updates.md |
| PostgreSQL (prod) | Production DB | Compute hours (suspend after 5 min idle) + max monthly storage | $0.16/compute-h; $0.35/GiB-mo (10 GiB cap); dev DB free (20 GiB/app) | same, https://docs.replit.com/billing/about-usage-based-billing.md |
| App Storage | Object storage | GiB-month + operations | $0.015/GiB-mo; $0.0004/1k basic, $0.005/1k advanced ops | https://docs.replit.com/billing/aug-cloud-billing-updates.md |
| Agent (effort-based) | Using Replit Agent | Per checkpoint, scaled by task effort, at model-provider API rates, from credits; daily Free Mode allowance on Core/Pro | no fixed $/h | https://docs.replit.com/billing/ai-billing.md, https://replit.com/pricing |
| Dev workspace | Building in the IDE | Not metered ("Development Time: Unlimited"); resources per plan unpublished officially | **Third-party only**: Core 4 vCPU / 8 GiB, Starter 2 GiB RAM | https://replit.com/pricing, https://www.nocode.mba/articles/replit-pricing |
| Pre-Aug-2026 prices (historical) | Billing cycles before 2026-08-01 | Old rates | CU $3.20/1M, req $1.20/1M, egress $0.10/GiB, dedicated 4/16 $0.22/h, 2/8 $0.11/h | https://docs.replit.com/billing/aug-cloud-billing-updates.md |

Regions: North America, Europe, Asia, South America, Australia (Core/Pro/Enterprise; permanent per published project;
no regional price difference published).

## Gotchas

1. **Not a sandbox API**: one Reserved VM = one published app. Running 50 parallel agent machines means 50 deployments;
   the account hard limit is 20 concurrent Replit Apps (workspace side), and there is no SDK to spawn VMs.
2. **Fee = credits, but credits are shared with Agent**: heavy Agent use can drain the credit before hosting does.
   Core credits reset monthly; Pro credits roll over two months; packs expire after 6 months.
3. **Reserved VMs are always-on**: listed per month; stopping between shifts to save money is not documented.
4. **Autoscale is cheap per vCPU-hour but request-bound**: billed only while serving HTTP requests; no background work
   between requests, and the machine size menu is unpublished.
5. **Autoscale base fee inconsistency** ($2 vs $1) and "legacy" shared 1 vCPU/4 GiB Reserved VM *rose* in Aug 2026.
6. **Budgets are hard stops**: hitting the budget suspends all usage-based services (your deployments go down).
7. **Geography is permanent** per published project; Starter is North America only.
8. Workspace vCPU/RAM per plan only appear on third-party sites now.

## Worked example

4 vCPU / 8 GiB, 50 concurrent × 8 h/day × 22 days = 8,800 machine-hours, 30% CPU, 50 GiB snapshots, 100 GiB egress.

| Regime | Feasible? | Monthly total |
|---|---|---|
| Reserved VM 4/16 (smallest ≥ 4 vCPU & ≥ 8 GiB), 50 always-on, Core | Yes (50 deployments) | 50 × $130 = **$6,500** (credit $20 absorbed; engine at 730 h: $6,591.90). Egress 100 GiB within the 100 GiB included → $0. No snapshot product; 50 GiB in App Storage would add $0.75 |
| Same on Pro $100 | Yes | $6,500 (fee fully consumed as credit; buys rollover, 15 builders) |
| Reserved VM only during shifts (if unpublishing daily worked; undocumented) | Unverified | 8,800 × $0.1806 = $1,589.28 |
| Autoscale 4 vCPU / 8 GiB, busy 100% of shift (alt) | Only for request-driven apps | 8,800 × $0.19008 = $1,672.70 + 50 × $2 base = **$1,772.70** + requests (30% CPU util saves nothing: billed on allocation while serving) |
| Autoscale, busy 30% of the shift | Request-driven | 2,640 h × $0.19008 = $501.81 + $100 base = $601.81 |
| Starter | No (1 app, expires in 30 days) | n/a |