The Nightglass story

The on-call desk that opens the fix.

Nightglass was built around a single observation: alerting is no longer the bottleneck — the stretch from page to fix is. We are building the desk that closes that stretch, so the rotation reviews a change instead of writing one under duress.

SOC 2 Type II
ISO 27001
HIPAA · BAA
EU residency
Why an autonomous on-call desk

Alerting is solved. The stretch from page to fix is not.

Modern alerting is fast. Alert correlation layers can fuse logs, traces, and deploy markers into a single incident card in seconds — and then escalate. The on-call still has to wake up, read the diff, decide whether to revert, write the patch, and queue the rollback. The correlation layer has handed the page off, but the work has just started.

The bottleneck on a bad night is no longer detection; it is the stretch from page to fix. Even an engineer who rolls out of bed in two minutes needs another ten to read the failing path, decide what changed in the last hour of commits, and produce a patch that will hold under review. That is a quarter-hour of customer-impact window multiplied by every incident the rotation absorbs in a quarter.

Nightglass pulls the fix forward into the desk. The same desk that correlates the signal drafts the PR, embeds the rollback, queues the Slack thread, and stages the one-click revert behind the merge button. The on-call still reviews and still merges, but they review a change instead of writing one.

What we are building toward

Every incident arrives with a fix queued and a rollback wired.

The founding mission is to make every incident arrive with a fix queued and a rollback wired. Not a recommendation, not a draft the on-call has to edit — a PR that the on-call can read, merge, or revert in under a minute. The desk writes the postmortem. The on-call signs off.

Nightglass is being shaped into a colleague, not another alert-correlation layer. That means reading the repo, learning the team’s conventions, calibrating confidence against the team’s error budget rather than a generic noise threshold, and routing every action through the channels the team already lives in. The work is to give every operation a desk that already knows its codebase.

The near-term arc is wider integrations and tighter confidence. Wider on alert sources and Git hosts so Nightglass fits any pager rotation; tighter on the PR-confidence floor so the on-call desk only opens fixes that will hold under review. The end state is an on-call rotation that reviews code instead of writing it under duress.

What the 3am pager experience actually looks like

From the page to a queued PR in under fifteen seconds.

The same artefact the home console renders, told as a sequence. Each step is one beat of an incident that opens against checkout-api at three in the morning.

  1. 01[detect]

    SLO breach fires at INC-2419 on checkout-api. error_budget_burn = 4.2× against payment.charge, severity scored against the team’s budget — not a generic noise floor.

  2. 02[correlate]

    The desk reads the last hour of commits on checkout-api and matches the failing path to feat/checkout-retry-with-jitter (commit a91f3c, deployed 38 minutes ago).

  3. 03[triage]

    SLO at 38% of budget. Severity scores SEV-2; ownership maps to the checkout rotation. Page routes to #sre-checkout with Joon on call.

  4. 04[open]

    The desk opens PR #1841 against checkout-api with the rollback pre-wired — revert queued behind the merge. A Slack draft posts the same channel; the on-call reviews the diff and picks merge or revert in one click.

Who builds the on-call desk

Operators who would rather write code than page humans.

A small desk. Engineers and reviewers, not managers. The same people who carried a pager rotation built Nightglass — and they still answer their own pages on every release.

Engineering

the on-call desk lead

Owns the auto-PR path from SLO breach to revert.

Reviewer and rotation owner. Watches every patch the desk opens against the team’s error budget, signs off on the queue, and writes the postmortem when something reaches prod. Writes fewer PRs than they review — on purpose.

SRE

the rotation PM

Runs the pager rotations and the desks that write their postmortems.

Pager veteran. Keeps the desk wired into every channel the rotation already lives in — alert sources, traces, PR queues, chat threads. Holds the bar on the cron: every review ships under fifteen seconds, and the dashboards prove it quarter over quarter.

Platform

the integration engineer

Holds the alert sources, Git hosts, and chat channels the desk talks to.

Maintains every webhook the desk listens on and every Git host it opens PRs against. The reason an alert and the queued PR line up is this person tracking both sides of the bridge — and the SLI that measures whether the fix actually stuck.

Product

the postmortem editor

Edits what the desk writes so the on-call reads prose, not stutters.

Reads every incident card, queued PR description, and postmortem the desk ships. Catches the tone slips that survive a 3am write and turns them into prose the on-call can review without coffee — by the next incident, every time.

Bring Nightglass into the rotation

The desk is ready when the rotation is.

Pick a plan and start checkout, or book a working session with the team that builds it.

next step

Pick a plan / request a demo.

Three tiers, one product. A real engineer reads every demo request.

See pricingRequest a demo