US-EAST · EU-CENTRAL · US-WEST · AP-SOUTH

Pick your cores and region.
We handle the rest.

Launch web apps, managed PostgreSQL, AI agent sandboxes, cloud desktops, and GitHub Actions runners in under a second. Pay only for active CPU and memory — strictly $0/hr when sleeping.

QUICK LAUNCH CONFIGURATOR
INSTANT PROVISION · ACTIVE-USAGE BILLING
AI Sandbox
20 GB NVMe included
Virginia, US
Active Rate
$0.0936/hr
Idle / Sleeping
$0.0000/hr
Selected Profile
2 vCPU · 4 GB · us-east
$ uncage create sandbox --vcpu 2 --ram 4g --region us-east
Launch Now
< 200ms
Cold Boot & Live Fork
Instant isolated VMs and copy-on-write forks
$0.00/hr
Idle Compute Rate
Zero charge while waiting on LLM tokens or traffic
4 Regions
Global Edge Footprint
us-east · eu-central · us-west · ap-south
Up to 85%
Lower Monthly Spend
Compared to always-on cloud VMs & managed PaaS
AUTOMATE EVERYTHING

Built for scripts, SDKs, and MCP agents.

Every workload shares the same predictable REST API and lifecycle commands: create, exec, pause, resume, and fork.

Zero vendor lock-in or capacity reservations
Consistent endpoints across all 4 global regions
Open internet egress unlocked with $5 verified balance
# Launch an isolated AI sandbox in us-east
$ uncage create sandbox --name swe-agent --vcpu 2 --ram 4g --region us-east
✔ swe-agent running  (2 vCPU · 4 GB · us-east)  https://swe-agent.sbx.uncage.dev

# Execute code directly inside the running environment
$ uncage exec swe-agent -- pytest -q tests/
14 passed in 0.82s

# Fork memory + disk state instantaneously for parallel evaluation
$ uncage fork swe-agent --name swe-agent-branch-b
✔ swe-agent-branch-b running in 180ms

# Pause to $0.00/hr compute when idle
$ uncage pause swe-agent
ACTIVE-USAGE PRICING

Pay when cores work. Zero when they sleep.

CPU $0.036/vCPU-hr · RAM $0.0054/GB-hr · Disk $0.10/GB-mo
Open Interactive Savings Calculator →
Compute Spec NVMe Disk Active Rate Sleeping / Idle Action
1 vCPU · 1 GB 10 GB included $0.0414/hr $0.0000/hr Launch →
2 vCPU · 4 GB 20 GB included $0.0936/hr $0.0000/hr Launch →
4 vCPU · 8 GB 40 GB included $0.1872/hr $0.0000/hr Launch →
8 vCPU · 16 GB 80 GB included $0.3744/hr $0.0000/hr Launch →
PLATFORM FAQ

Common questions from engineering teams.

01 What workloads can I run on uncage.sh?
uncage.sh unifies five core developer workloads under one CLI, SDK, and rate card: HTTPS Web Apps, Managed Databases (PostgreSQL 17, MySQL 8.4, Redis 7), AI Agent Sandboxes, browser-streamed Cloud Desktops, and drop-in GitHub Actions CI Runners.
02 How does $0/hr idle billing actually work?
When a sandbox, preview app, or desktop goes idle—or when you call `uncage pause`—compute metering stops immediately. You pay $0.00/hr for CPU and RAM while paused, and only $0.10/GB-mo for stored disk blocks. The workload wakes in under a second when a request or command arrives.
03 How fast are live memory forks and database branches?
Live sandbox memory forks complete in ~180ms using copy-on-write snapshots so AI agents can branch and evaluate multiple code paths in parallel. Managed PostgreSQL branches clone multi-gigabyte databases in under a second without duplicating unchanged storage blocks.
04 Can autonomous agents like Claude or Cursor control uncage.sh directly?
Yes. Install `@uncage/mcp-server` in your MCP client config to expose sandbox creation, command execution, live forking, database branching, and cloud desktop control directly to Claude Desktop, Cursor, or custom agent loops.
START SHIPPING IN 30 SECONDS

Pick your cores and region. Pay $0 when idle.

Every new workspace starts with $10 in starter credit. Launch web apps, managed PostgreSQL, AI sandboxes, cloud desktops, and CI runners with zero lock-in.

✓ $10 starter credit on signup
✓ Strictly $0.00/hr compute when sleeping
✓ Sub-second boot across 4 global regions