AgentSafe – per-task micro-VM sandbox for AI agents (Go)
AgentSafe – per-task micro-VM sandbox for AI agents (Go)
I was tinkering with AI agents for some time and made some utilities for myself for safe and fast execution. While chatting, I figured I can open source it. Spent some time cleaning and vibe coding some extensions for it. Here the results from real execution on M1 mac: VM Boot Times: /tmp/vm_final_1.log:VM Boot Time: 0s (Target: <200ms) /tmp/vm_final_2.log:VM Boot Time: 1ms (Target: <200ms) /tmp/vm_final_3.log:VM Boot Time: 0s (Target: <200ms)
Share cardActual performance
Launch Intel predictions
Analyze your own launch →Correct prediction on native model
Similar products
A1 – Local Sandbox and JIT Compiler for AI Agents
Amla Sandbox – WASM bash shell sandbox for AI agents
Secure sandbox for AI Agents. 100% POSIX.
Azure-sandbox
Helios OpenTelemetry Sandbox
ECMAScript Sandbox
ECMAScript Sandbox [WIP]
Spacedeck Sandbox
SandBox – AI agents simulating possible futures
Instant, hardware Isolated Sandboxes for AI agents