When users are tired,
AI should slow down.

Humsana adds real-time cognitive awareness to AI tools — so risky actions don't happen at the worst possible moment.

Install in minutes

pip install humsana-daemon
npx @humsana/mcp-server
The Problem

Developers don't make their worst mistakes randomly.

They make them when they're sleep-deprived, cognitively overloaded, context-switching, or rushing late at night. AI systems can't see this.

When You're Impaired

  • Sleep-deprived
  • Cognitively overloaded
  • Context-switching
  • Rushing late at night

AI Does This Anyway

  • Executes destructive commands
  • Overwrites large files
  • Makes irreversible changes
  • Assists with false confidence
The Solution

Cognitive Interlock

Humsana introduces a cognitive safety interlock between users and AI execution.

Think seatbelts, not autopilot.

When cognitive strain is detected:

Dangerous commands are blocked
Large destructive writes are paused
AI responses adapt automatically
Overrides require deliberate confirmation
How It Works

Four steps to cognitive safety

01

Local daemon observes behavioral timing

Lightweight process monitors typing rhythm, corrections, and activity — never content.

02

Signals analyzed locally in real-time

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

03

AI tools query state via MCP

Claude, Cursor, or custom agents get structured context about how you're doing.

04

Unsafe actions gated when needed

Destructive operations blocked. Overrides logged. Audit trail preserved.

Design Principles

Detection is passive
Blocking is intentional
Privacy is non-negotiable
Latency is single-digit ms
MCP Tools

What AI Receives

Four tools available to Claude and other MCP-compatible agents:

humsana:get_user_state

Returns stress_level, focus_level, fatigue_level, uptime_hours, and response style recommendations.

humsana:check_dangerous_command

Checks if a command matches dangerous patterns and whether user fatigue exceeds threshold.

humsana:safe_execute_command

Executes shell commands with interlock protection. Blocks when fatigued, logs overrides.

humsana:safe_write_file

Blocks large destructive rewrites (>30 lines removed) when user is fatigued. Saves to review folder.

Example response from get_user_state:

{ "state": "focused", "state_label": "🎯 Deep Focus", "metrics": { "stress_level": 0.30, "focus_level": 0.72, "cognitive_load": 0.55, "typing_wpm": 45, "fatigue_level": 28, "uptime_hours": 3.2 }, "fatigue": { "level": 28, "category": "low", "uptime_hours": 3.2 }, "recommendations": { "style": "direct", "length": "concise", "ask_clarifying_questions": false, "tone": "efficient, don't interrupt flow" } }
Protected Commands

What Gets Blocked

The interlock activates when dangerous commands meet high fatigue:

rm -rf and sudo rm
DROP TABLE and DROP DATABASE
git push --force
kubectl delete
terraform destroy
docker system prune

Override Protocol

When Humsana blocks a risky action, the user can still proceed — but only by explicitly declaring intent.

OVERRIDE SAFETY PROTOCOL: [reason]
Forces deliberate acknowledgment
Prevents accidental execution
Creates an audit trail
Preserves user autonomy

Overrides are logged to ~/.humsana/audit.log for post-mortems.

Privacy

Privacy Model

Trust is not optional.

No keystrokes captured
No screen inspection
No cloud sync by default
Open-source daemon
User-controlled data
Fully auditable code

Start free. Upgrade when you need interlocks.

Free

$0
  • Local daemon
  • MCP integration
  • Adaptive AI responses
  • Fully on-device processing
Get Started