The mind lives outside the app.By design.
The app is data and screens. The mind is an external agent that reads the system through MCP and writes back. You can swap it, switch it off, or ask exactly what it did.
Agent
thinks on a schedule
MCP
tools with boundaries
Your database
source of truth
zero LLM in the app code — the mind is external and replaceable
Rituals of the day
The agent works on a cadence — it doesn’t “always listen”, it shows up at its hours.
Every morning · 05:30
Brief and day plan
Reads sleep, calendar, and tasks, then assembles the day before you wake: focus blocks placed, conflicts flagged.
intelekt · agent
05:30
Every evening
Day review
Closes the loops: what got done, what moves and why. Notes people from your messengers — touchpoints land on the timeline by themselves.
Day review
closed 6/7
Focus: 2 h 10 min of deep work
“Set up auth” → moved to tomorrow
Marta: call note added
Every week
Insights across areas
Looks at weeks, not days: trends, correlations, quiet drifts. The things you never notice inside the daily routine.
Insight of the week
Sleep under 7 h → next-day mood drops by 1.8 points.
based on 21 days
Honest by design
The agent says “worth noticing”, it doesn’t pass verdicts. Every record it writes is signed and visible: what it did, when, and on which data.
What the mind sees
- Tasks, projects, and the calendar
- Health aggregates: sleep, readiness, workouts
- Financial totals — no account numbers
- Notes and journal outside the vault
What it never sees
- Vault records — outside auto-context
- Passwords, documents, card numbers
- Anything marked private — even in synthesis
- Anything you didn’t explicitly grant
The vault isn’t a setting — it’s an architectural boundary: private data technically never enters the agent’s context.