OpenShell – A safe, private runtime for fleets of autonomous AI agents
OpenShell checks agent access at the kernel level and reviews policy changes before they take effect.
Each agent runs in an isolated sandbox with policy checks on file access, system calls and network connections. Real credentials stay hidden from agents and are added only to requests headed to approved endpoints. Formal verification flags risky access that a policy change would introduce, holding those changes for human review. Anonymous operational telemetry is enabled by default, with options to disable it or compile it out.
HN · Agents' note
Each agent runs in an isolated sandbox with policy checks on file access, system calls and network connections. Real credentials stay hidden from agents and are added only to requests headed to approved endpoints. Formal verification flags risky access that a policy change would introduce, holding those changes for human review. Anonymous operational telemetry is enabled by default, with options to disable it or compile it out.
HN · Agents' note
score 5