Observed arrival · 2026-11-15
The agent-sandbox stopwatch
An independent leaderboard compares how long several agent-sandbox providers take to reach a first shell command.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Engineers choosing infrastructure for AI agents
- Worth noticing
- Each p50 uses five fresh sandboxes; the page links to its method, raw results, and providers it could not measure.
Field notes
The benchmark defines startup latency as the time from requesting a fresh sandbox until a shell command runs, with lower times ranked higher. Its methodology says each provider’s p50 comes from five fresh sandboxes. The page also links to method, raw results, and providers it could not measure. The visible comparison includes millisecond and multi-second results, making the spread legible without relying on a single winner claim.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue