Persistent workspaces.
Built for agents.
Long-running, isolated workspaces with persistent disk for agentic and MCP workloads. Pause when idle, resume in milliseconds — never pay for a sleeping agent.
1from asc import Workspace 2 3# Create or attach to a persistent workspace 4ws = await Workspace.from_id( 5 "user-42", 6 cpu=1, mem="2Gi", 7 disk="20Gi", # persists across hibernations 8) 9 10await ws.fs.write("/work/notes.md", "Hello, agent!")11print(await ws.exec("ls /work").stdout)12 13# Idle hibernation in 60s — billing pauses automatically.Designed for production.
Hibernate + wake
Idle agents hibernate within seconds; wake on the next message with full state intact.
Persistent volumes
Each workspace gets its own disk that survives restarts, migrations and code updates.
Shell, files, browser
Built-in tools: PTY shell, filesystem API, headless browser, jupyter — all sandboxed.
MCP-native
Workspaces speak Model Context Protocol out of the box. Plug into Claude, Cursor, ChatGPT.
Per-tenant isolation
Network policies and resource quotas enforced per workspace — safe for multi-user agents.
Observability
Stream every tool call, command and HTTP request to your own OpenTelemetry collector.
Metered. No markup.
Pay per active second / per GiB. Free tier covers small projects; $200/mo cap until you opt in. See the full calculator.
| Line item | Unit | Rate (USD) |
|---|---|---|
| Workspace 1 vCPU | per hour (active) | $0.046 |
| Sandbox 0.25 vCPU | per hour (active) | $0.014 |
| Persistent storage | per GiB-month | $0.026 |
| Egress | per GiB | $0.098 |
Ship your first deploy in minutes.
Free $30/month of compute. No card required.