Policy
AI Firewall
Decide what your AI is allowed to do. Observe softly or hard-block dangerous actions.
Observe
1 rule
Passive logging, no interruptions.
Warn
0 rules
Surface a notification before the action.
Approve
1 rule
Require explicit user approval first.
Block
2 rules
Hard-stop the action and log the attempt.
Warn on credential file accessapprove
Detect access to secrets folders
file_read MATCHES /\.aws|\.ssh|credentials/
Require approval for medical promptsblock
Healthcare safety guardrail
prompt CONTAINS medical|diagnose|prescribe
Block shell exec on autonomous agentsblock
Prevent agents from running shell commands
event_type=tool_exec AND process_kind=autonomous-agent
Observe local LLM activityobserve
Passive logging only
kind=local-llm