DevelopersCognitive Interlock

When you are tired, AI should slow down.

A safety interlock between you and AI execution. Think seatbelts, not autopilot.

pip install humsana-daemon npx @humsana/mcp-server
01The problem

The worst mistakes are not random.

They happen when you are sleep-deprived, overloaded, or rushing late at night. AI cannot see that, so it executes destructive commands and overwrites files with false confidence anyway.

01

Observe timing

A local daemon monitors typing rhythm, corrections, and activity. Never content.

02

Analyze on-device

Stress, focus, and fatigue are computed on your machine. No cloud required.

03

Gate unsafe actions

Destructive operations are blocked when fatigue is high. Overrides are logged to an audit trail.

02MCP tools

What Claude and other agents receive.

get_user_state

Stress, focus, and fatigue levels, plus response-style recommendations.

check_dangerous_command

Whether a command is dangerous and fatigue exceeds threshold.

safe_execute_command

Runs shell commands with interlock protection. Blocks when fatigued.

safe_write_file

Blocks large destructive rewrites when the user is fatigued.

Blocked when fatigue is high: rm -rf, DROP TABLE, git push --force, kubectl delete, terraform destroy. Override with OVERRIDE SAFETY PROTOCOL: [reason], logged to ~/.humsana/audit.log.
03Pricing

Start free. Upgrade for interlocks.

Free
$0
  • Local daemon
  • MCP integration
  • Adaptive AI responses
  • Fully on-device
Get started
Pro
$10 / mo
  • Everything in Free
  • Cognitive Interlock
  • AI write protection
  • Override audit logs
Start Pro