Local-first · reads ClickUp, Obsidian & git

The chief of staff who already read everything before you woke up.

You run several ventures on a two-hour morning window. Before it opens, Zorapilot reads your ClickUp, your Obsidian vault, and every repo, then hands you one brief. Once you trust that read, it starts clearing the small things itself.

It starts by reading. Then it earns the right to act.

The re-orientation tax

With many ventures, the scarce resource isn't time. It's context-switching capacity.

Every morning opens with the same tax: wait, where is everything? You pay it across ClickUp, a stack of repos, and an Obsidian vault before a single decision gets made. And the real failure mode isn't bad calls, it's a project silently dying in the corner while nothing flags it. Zorapilot collapses that triage into something already written by the time you sit down.

Stage one · the read (start here)

It reads your whole operation, then hands you one brief.

This is the low-trust way in. Read-only, pennies a run, nothing changes but how much you already know when you sit down. Three standing jobs keep ClickUp, your vault, and your repos in one view.

Project Sync · nightly

Dashboards stop lying

Reads git and ClickUp, then writes a machine-owned State.md next to every project and raises drift flags when reality and the dashboard disagree.

# State: New Website
Status:      quiet
Last code:   no commits in 70d
Drift: says "active build" · 0 commits
Daily Brief · weekdays

Start on execution

The work window opens with today's priorities from ClickUp, what changed overnight in your repos, and what's gone quiet, not a blank inbox.

Today's priorities
1. Safety doc · 2d overdue
2. Ship round-timeline
Changed: 30 commits, 5 blockers closed
Sunday Packet · weekly

Planning day, pre-done

The weekly review is written before you open it: what shipped, what's at risk, and a suggested top three, so Sunday is react-and-decide.

Week ahead
Suggested top 3
· Close the launch blockers
· Kill or restart the frozen GTM
· Ship the scheduler

The insight

A human can't watch fourteen things for silence. A machine can.

The projects that fail don't blow up, they go quiet. A mechanical no movement in 14 days check every morning fixes an attention limit you were never going to beat by trying harder.

drift flags · this morning
Capture-the-Flag GTM · 92d silent, 84d overdue
Advisory Council · repo never initialized
Website · "active", 0 commits/7d

Stage two · the act (the point)

The brief is the wedge. Doing the work is the point.

It doesn't stop at telling you. Every job it runs is already a governed write. It maintains a State.md next to each project and files your briefings into your own vault, inside a scope you set, logged and reversible. That's real action on day one. And because every action rides the same rails, widening what it may touch is a setting you loosen as trust builds, not a new product to buy.

Today · shipped

It writes, not just warns

State files and briefings land in your vault as durable, dated artifacts, each write checked against the scope you granted before it happens.

write _System/Briefings/2026-07-11.md
write Dodgeball/_Projects/…/State.md
ok in scope · logged
The rails · already built

Scope is a dial, not a rebuild

Roles, write scopes, and cost caps live in settings. Grant a role a wider path and its next run acts there. Deny it and the attempt is blocked and logged.

role: syncer  cap: $2/run
writeScope: _System/Briefings/
deny write outside scope → logged
Where it's headed

From clearing to closing

The same governance that lets it write your vault is what will let it close the tasks you already triaged and move work forward, one scoped, reversible step at a time.

Once you widen scope
· close what you marked done
· file the loose note
· nudge the blocker

Governance, not ceremony

The difference between a demo you babysit and infrastructure you trust while asleep.

These are the rails that make stage two safe. They're why letting it act is a setting you loosen, not a leap of faith.

Scoped writes

Every agent can only write where you allow. A write outside its scope is denied and logged, not attempted.

append-only audit log

Hard cost caps

A per-run spend ceiling halts a runaway before it burns money. You see cost on every run.

you set the ceiling · ~$29 / mo full schedule

Approval pauses

Anything sensitive escalates to you and waits. No answer in time defaults to blocked, never to "went ahead."

timeout → block

Your machine, your keys

Local-first: it installs on your machine and reads your own vault directly. Bring your own Anthropic key.

nothing leaves the box

Straight talk

Two things about it, said out loud.

It only pays if you read the briefs.

An unread brief is a diary that writes itself. The leverage is real, but it starts the moment you actually open the thing each morning.

Today it reads and writes your vault. Closing ClickUp tasks is where it's going.

The shipped action is durable artifacts and drift flags in your own vault, under governance. The rails to reach further are built. It earns the wider scope by proving the read first.

Deployment #1

The case study is a real operator's morning.

Zorapilot runs a solo founder's four ventures every day, on a launchd schedule that survives a reboot, for the price of a couple of coffees a month. If it demonstrably works on one busy life, that's the whole pitch.

4
ventures, one operator
~$29
per month, full schedule
$0.006
per daily brief
3
jobs, every morning

Get started

Wake up to a desk that's already been cleared.

One guided install sets up the app and the background scheduler. Point it at your Obsidian vault and connect ClickUp for the full read, bring your Anthropic key, and let it take the first shift. Only run on repos and a vault? It works on those alone.

$19/ month

one plan · bring your own Anthropic key · cancel anytime