Code executionOnline
Daytona
Isolated sandboxes for AI-generated code: Linux containers by default, plus VM, GPU, Windows and macOS runtimes.
- Cold start
- 90 ms
- Compute rate
- 504 $/10k vCPU-h
AgentSandboxes lists 25 sandbox providers with source-linked facts from each vendor's own docs.
Code executionOnline
Isolated sandboxes for AI-generated code: Linux containers by default, plus VM, GPU, Windows and macOS runtimes.
Code executionOnline
Firecracker microVM sandboxes for agent-generated code, with pause/resume and filesystem snapshots.
Dev environmentsOnline
Full Linux VMs for agents: 65ms boot, live forking, pause/resume and persistent snapshots.
Code executionOnline
Bedrock agent runtime where each session gets a dedicated microVM; per-second billing, plus code-interpreter and browser tools.
Dev environmentsOnline
Perpetual microVMs that suspend to zero when idle and resume in about 25ms with filesystem and memory state intact.
Browser sessionsOnline
Hosted headless-browser sessions for web agents, with the MIT-licensed Stagehand SDK.
Six mistakes that teams make when picking a cloud sandbox provider and the questions to ask before you commit, based on specs recorded in this index.
Compare sandbox providers by isolation, cold start, persistence, GPU support, pricing model and free tier using a standard checklist and the compare tool on this site.
Every listing on this site records the specs a provider publishes on its own docs or pricing page: isolation, cold start, persistence, GPU, pricing and more. Each row links the cited source.
Every provider in this index serves the Online or Global scope. This guide covers what that means for latency, data residency, team access and compliance.
Published by AgentSandboxes. How we check each sandbox provider.
A short email now and again: a new provider, a pricing change that matters, and one source note. No noise.