One glance, not a scrollback
The Active-Idle-Done board shows every session at once: project, branch, source badge, live metrics. Read the room in two seconds, not from terminal scrollback.
macOS menu-bar app for long agent runs
Stop babysitting your coding agent.
No more staring at a terminal or carrying an open laptop around just because Claude Code is still running. Flip on Safe Carry, close the lid, walk away — come back to a receipt of what it did.
Covered while you step away — not magic against a dead battery. macOS 14+.
Runs stay alive lid-shut Active / Idle / Done board Rollback notes in every receipt Local-first, no account $9 once, no subscription
Sessions 5 sessions across all sources
CLI 2Codex 1VS Code 1Desktop 1
Active 2
auth-hardening-42 CLI
▦ 24 prompts ⇄ 61 turns ◷ 38m
permissive ⚡ Skip-perms
payments-webhook Codex
▦ 12 prompts ⇄ 30 turns ◷ 14m
strict ⚡ Skip-perms
Idle 1
docs-site VS Code
▦ 6 prompts ⇄ 11 turns ◷ 5m
manual ⚡ Skip-perms
Done 2
rate-limiter Desktop
▦ 31 prompts ⇄ 74 turns ◷ 52m
permissive ⚡ Skip-perms
WatchingSafe Carry on · 5 sessions
What it does
You don't watch terminals. You run them.
One agent or four across three repos — you shouldn't hover over a terminal to see what happened. Claude Watcher gives you the board, keepalive, and receipt so you can step away.
Close the lid. The run keeps going.
Safe Carry blocks closed-lid sleep so Claude Code or Codex keeps running while you take a call, commute, or move rooms. Covered while you step away — not pinned open on your desk.
Come back to a receipt, not a guess
The after-action receipt lists changed files, token burn, risk labels (secret-like paths, destructive commands, force-push blocked), and the exact files to restore. Trust it because you can audit it.
See it run
The first thing you check when you sit back down.
Not a log to scroll — a receipt. One run summarized: runtime, token burn, changed files, risks flagged, and the exact files to restore. Read it in ten seconds, then decide if you trust it.
Safe Carry, safe auto-approve, and after-action receipts are part of Solo Lifetime ($9). The board, Today, and Trace are free.
How it works
Walk away safe, not reckless.
01
Start the run, hit Safe Carry.
Kick off Claude Code, Codex, or a long shell job the way you already do. Tap Safe Carry and Claude Watcher blocks closed-lid sleep, so the run keeps going when you shut the laptop or step away.
safe carry02
It watches so you don't have to.
The board tracks every session across Active, Idle, and Done. Low-risk prompts move on a policy you set; risky actions get held, not waved through. Force-push gets blocked. Secret-like paths get flagged.
sessions board03
Come back to a receipt you can trust.
Open the receipt: changed files, token burn, risk labels, and the exact files to restore. Trust the run, or undo it by hand. You stay in control either way.
receiptWhat you get
What you actually get.
Real features, not a wishlist. Each answers one question: what did it do while I was gone, and can I trust it? Local-first — your code never leaves your Mac.
After-action receipt
A readable summary for every run: time away, changed files, token burn, risk labels, and result. The thing you read first when you sit back down, instead of scrolling a terminal to piece it together.
Risk labels
It flags the stuff you'd want to catch: secret-like paths (.env, key files), destructive commands, and force-pushes blocked before they land. Not a vibe. A labeled list you can scan.
Rollback notes
When a run goes too far, you get the exact files to restore and which changes to undo. No silent auto-rewrite of your repo. You decide, with the receipt in front of you.
Active / Idle / Done board
A kanban for your sessions. Each card shows project, source badge (CLI / VS Code / Desktop / Codex), short session id, git branch, and metrics: prompts, turns, duration, last activity. Many runs, one glance.
Safe Carry keepalive
Blocks closed-lid sleep so a run keeps going while you close the laptop, take a call, or move rooms. Covered while you step away. It can't survive a hard shutdown or a dead battery, but nothing can.
Safe auto-approve
Low-risk Claude Code prompts move automatically; risky actions are held for you. Set a per-session policy — Global, Permissive, Strict, or Manual — right on the card. A safe policy plus a record, not a skip-all bypass.
Trace lens
Token analytics per session: input, output, cache-read, cache-write, and cache-hit-rate (higher is cheaper), with model badges. Full-text search across prompts, tools, and output, plus adjacent-turn diffs.
Today dashboard
Today's prompts, context switches, active projects, and total sessions. Today-by-project bars, 7-day task clusters, a histogram, and a weekly report generated locally with Ollama (rule-based fallback if it's off).
Prompt coaching
Optional and local. When a Claude Code prompt looks vague — no path, no action verb, no number — it suggests a sharper rewrite and drops it on your clipboard, catching weak prompts before they burn a run. Runs through your own Ollama.
Pricing
$0 to watch. $9 to stop babysitting.
Free gets you the board and live status today. Solo is the impulse buy — the moment you want to close your laptop and trust what you come back to. No subscription, no renewal email later.
Free
$0
See the board
- 7 days of full Solo to start — free, no card
- Sessions board: Active / Idle / Done
- Today dashboard + Trace token view
- Live status across CLI, VS Code, Codex
- Local-first, no account, no telemetry
Solo Lifetime
$9
Pay once · lifetime
- Everything in free
- Safe Carry keepalive — close the lid
- Safe auto-approve, per-session policy
- After-action receipts + rollback notes
- Local prompt coaching (Ollama)
- Every future update, license stays local
Payments by Lemon Squeezy. One-time $9 — a lifetime license verified locally with Ed25519, no online activation, no account. Not for you? Email within 14 days for a refund.
FAQ
Small questions before checkout.
Does my agent really keep running when I close the laptop? +
Yes, within reason. Safe Carry blocks closed-lid sleep so runs keep going while you step away or carry your laptop. It doesn't survive a hard shutdown, crash, or power loss — nothing on the Mac does. It keeps the machine awake, not invincible. Think “covered while you walk to the kitchen,” not “survives anything.”
What's actually in the after-action receipt? +
Changed files, token burn, risk labels (secret-like paths, destructive commands, force-push blocked), and rollback notes naming the exact files to restore. It lets you trust a run you didn't watch — in ten seconds, not ten minutes of scrollback.
Which agents and sources does it watch? +
Claude Code, Codex, and long terminal/CLI agents. The board tags each session with a source badge (CLI, VS Code, Desktop, or Codex) and tracks many sessions across many sources at once. Automatic live triggering and safe auto-approve are Claude Code-first today; deeper Codex tracking is on the roadmap.
Is Safe auto-approve just 'skip all permissions'? +
It isn't a blanket bypass — that's the whole point. Low-risk Claude Code prompts can move automatically while risky actions are held. You set a per-session policy: Global, Permissive, Strict, or Manual. It's a safe policy plus a record you can audit.
Does it roll back my code automatically? +
It doesn't, by design. You get rollback notes — the exact files to restore and which changes to undo — and you decide. No silent rewrite of your repo behind your back.
Is my code or my prompts sent anywhere? +
No. Everything is local-first and your code never leaves the Mac. License verification is local too: Ed25519 signatures, no online activation, no account. There's no telemetry shipping your work anywhere.
Is the download safe to open? +
Yes. The DMG is signed with a Developer ID certificate and notarized by Apple, so macOS opens it without the “unidentified developer” Gatekeeper warning. It's a free, local app — no account, no telemetry.
Do I need Ollama? +
Only for the optional bits: prompt coaching and the generated weekly report. Both run locally through Ollama, and the weekly report has a rule-based fallback if Ollama isn't around. The board, keepalive, and receipts don't need it.
What does the Trace lens give me over my agent's own logs? +
Token analytics you can read at a glance: input, output, cache-read, cache-write, and cache-hit-rate (higher is cheaper), plus per-session rollups with model badges, full-text search across prompts, tools, and output, and adjacent-turn diffs showing tools added or removed and token deltas.
Free vs Solo Lifetime — what's the difference? +
Free gets you the board, the Today dashboard, and the Trace token view, so you can prove it works. Solo Lifetime ($9, one-time) unlocks the walk-away loop: Safe Carry keepalive, safe auto-approve policies, after-action receipts with rollback notes, and local prompt coaching.
Is this affiliated with Anthropic or OpenAI? +
No. Claude Watcher is an independent tool. It works alongside Claude Code and Codex but isn't affiliated with or endorsed by either company. Claude is a trademark of Anthropic.
What does it run on, and is $9 really one payment? +
A Mac on macOS 14 or newer. Solo Lifetime is a single $9 payment: no subscription, no renewal, no account. Pay once, unlock forever. Future updates included.
Can I get Solo Lifetime for free? +
Yes. Post an honest take about Claude Watcher on Reddit, submit the link, and after a quick manual review we'll email you a Solo Lifetime license. We want your real opinion, good or bad, not a planted five-star.
The impulse
"I just want to close my laptop and trust what I come back to."