Run your agent's routines by name.
Routines are reusable behaviors for your private agent: briefs, drafts, reminders, memory, follow-ups, and workflows you can run by name. Start with 48 built-in. Your agent can suggest new routines as it learns your repeated requests — always with your approval.
The eight routines most people use in the first week.
Each one shows what it needs, when it runs, and whether it can read, draft, or act.
A short summary of today the moment you wake up — meetings, open loops, anything overdue.
Finds the few emails worth attention and drafts replies in your voice.
Reads the calendar invite plus the last thread with that person and surfaces what matters.
“What do I know about Aleksy?” — pulls from notes and prior threads before the meeting.
Captures the decision you just made, the reason, and what you ruled out.
“Where were we on X?” — pulls the last commit, doc edit, and message thread.
Open loops you owe a reply on, surfaced before they go cold.
What you shipped, slipped, decided — in your usual end-of-week format.
Boring on purpose. No surprises.
Every routine follows the same shape. Nothing leaves your messenger or your sources without you saying yes.
Browse by what they're for.
Cards show only the metadata that's non-obvious. No badge means: runs on request, read-only, no integrations needed.
Heartbeats
7 routinesDrafting
7 routinesMemory
6 routinesWork
9 routinesLearning
3 routinesTravel
4 routinesRecommends and drafts first. Books only when you confirm — never silently.
Health
3 routinesOptional, private, disconnect anytime. Health routines never share data with anyone but you.
Money
3 routinesRead-only by default. Anything that moves money is drafted first and only sent after you confirm.
Personal
6 routinesFrom a repeated ask to a saved routine.
If you ask for the same thing a few times, your agent offers to turn it into a reusable routine. You decide whether to save it, schedule it, or let it act on its own.
You repeat. It notices. You approve.
No silent surveillance, no “the AI guessed.” The agent shows you the pattern it saw, the routine it proposes, and exactly where it can act without asking again.
Every suggested routine is editable before you save it — the schedule, the sources, the boundary. Delete it any time.
And it suggests new ones when patterns repeat.
If you ask for the same thing a few times, your agent can turn it into a reusable routine. You approve what gets saved, scheduled, or automated.
Request a private beta invite