Armorer – A secure local control plane to sandbox AI agents in Docker
Docker sandboxing for agents when most runners just trust whatever the LLM outputs.

Kernel-enforced agent sandboxing that blocks .env access without container overhead.
Developers running AI coding agents locally
Firecracker · gVisor · Bubblewrap
Docker sandboxing for agents when most runners just trust whatever the LLM outputs.
Native macOS VMs with APFS snapshots beat Docker for agent isolation.
Defense-in-depth sandboxing for Claude agents: Podman + MITM proxy + seccomp + iptables.
TLA+ verified sandbox makes --dangerously-skip-permissions safe for Claude Code and other agents on macOS.
Sandboxes the build phase, not just runtime — most agent security tools ignore npm install scripts.
Full macOS isolation beats containers for agents needing GUI apps and native tools.