A first-pass on-call operator for Claude

The pager that stays quiet
unless it's a real fire.

Nightwatch reads an incoming alert and makes the call your on-call engineer would make in the first 90 seconds — then drafts the message. Built for the team without an AIOps budget, where the scarcest resource is uninterrupted sleep.

It decides — it never kicks the question back.  PAGE · TICKET · SUPPRESS · FLAG, every time.

Watch it decide

Watch it make the call, in real time.

No narration, no edits. Three alerts in, three different decisions out — the way it would stand your watch tonight.

Three alerts · three calls · 9:16
The problem

Alerts all night. Most of them are nothing.

One to three people share the pager. There's no NOC, no SRE on staff, no AIOps platform — just a phone that buzzes at 3am for a disk-space warning that always clears itself.

So you do the math every on-call does: read the alert, guess the blast radius, decide in 90 seconds whether it's worth waking up for. The decision isn't hard. The volume is what burns you out — and the night it actually matters is the night you've learned to swipe the buzz away.

90 sec
the triage call a human makes per alert
0
of those calls should require waking that human
What it does

Every signal leaves with one decision.

No "here's what I found, what do you want to do?" An operator decides and routes the work. You come back to a correct page — or a quiet night.

PAGE

A real fire. Wake a human now, with the blast radius and a first check already drafted.

TICKET

Real, but it can wait for daylight. Drafted with severity, impact, and a suggested owner.

SUPPRESS

Known noise or self-healing. Logged with the exact tripwire that would change its mind.

FLAG

Genuinely your call — rare. It holds a safe default and asks one specific question.

See it decide

Three real alerts. Three different calls.

Severity is computed from business impact, not the alert's own label — so a global 0.2% blip can outrank a "CRITICAL," and a "CRITICAL" can be safely muted.

Real outputs — produced by the folder running in Claude, not mockups.

Prometheus · 03:10 UTC
Disk usage on prod-db-1 at 88%. Severity: warning.
⚪ SUPPRESS
Known nightly log-rotation spike (normal ≤92%, clears by 03:30). Escalate-if: >92%, or not below 80% by 03:45, or free space <5%.
Known self-healing pattern within its window — suppress the human cost, keep the safety net.
Sentry · 14:30 local
500s on /settings/profile, ~3% of users, stable 20 min, workaround: refresh.
🟡 TICKET · SEV3
Non-core page, degraded not down, workaround confirmed, stable. Queue this week. (Check: any deploy in the last 30 min?)
No reason to pull someone off focus — a fix-in-daylight bug, and daylight is now.
Datadog · 02:40 UTC
503s on the API write path, 1.5% globally, stable — all from org_2290.
🟠 PAGE · SEV2 · SLA
1.5% globally = 100% of Acme Co.'s write path (Priority SLA, 4-hr clock running). Paging on-call + account owner.
Raw volume is small; business impact isn't. Severity is about who's down, not how many.
How it works

A minute to set up. Then it stands the watch.

Drop in the folder

Put the Nightwatch folder into a Claude Project. Claude becomes the operator — identity, rules, examples, and reference tables all loaded.

Teach it your world, once

Fill the reference/ templates with your services, severity thresholds, SLA customers, and recurring false alarms. ~30 minutes.

Paste any alert

It returns one decision, the routed action or drafted message, and a one-line "why" you can audit or override in five seconds.

The methodology

Folders as architecture. Each file does one job.

nightwatch/
├── identity.md      # who it is, what it refuses
├── rules.md         # the decision logic (the heart)
├── examples.md      # worked calls + edge cases
├── README.md        # how to use it
└── reference/
    ├── severity-matrix.md   # blast × impact × trend
    ├── routing-table.md     # sev → who, when
    ├── known-noise.md       # your false alarms
    ├── sla-customers.md     # who overrides the metrics
    └── response-templates.md# the drafted artifacts

The decision logic is a short-circuiting flow: security and data-loss gates fire first, then noise, then dedup, then severity from the matrix, then the SLA override. The first step that produces an outcome wins.

Nothing is a black box. The rules are plain English in a file you can read, edit, and trust — which is the whole point.

For builders

Want to build your own operator?

Nightwatch is one worked example. The Operator's Handbook is the method behind it — how to turn Claude into something that decides instead of chats.

The 5-file structure, and how to write decision logic with real thresholds — not "use good judgment"
How to handle the edge cases that break most tools, and keep the human gate where it matters
New operators as they ship from The Quiet AI — delivered with the handbook
Get the Handbook — free

A short, practical guide. No spam, unsubscribe anytime.

The folder itself is free and ungated — grab it on GitHub, no email required. This is just for the build-your-own guide.
No catch

It's a folder, not a SaaS.

Nightwatch is free. There's no account, no agent to install, no dashboard, no per-seat pricing. It's a folder of plain Markdown you drop into Claude — and it's running in a minute. Take it, fork it, make it yours.