Product
How it works, in detail
99.99% of events and incidents, answered by agents.*After a fine-tuning period of 30 to 60 days, depending on event volume and the integrations connected.
What we need from you
- One named approver, a platform admin or the CISO
- One escalation contact who can be reached at night
- Read access to the sources you choose: a least-privilege app or role per source
- A Slack workspace for the case rooms and the digest
At night
Urgent product support and the agents' work run at night; a person on your side answers the approvals. A 24x7 Managed SOC, our analysts taking what the agents hand to a person, is available on request.
Know what you own
Every laptop, account and cloud service your business runs on, who owns it and how much it matters.
Included in every NightWatch plan, built from Microsoft Intune, Microsoft Entra ID, Google Workspace, AWS and the service desk's asset register.
- Devices and their state from Microsoft Intune; accounts, sign-in methods and admin roles from Microsoft Entra ID and Google Workspace; accounts, keys and exposures from AWS.
- Owners and criticality from the service desk's asset register, so a case about a till, a scanner or the finance laptop is read with its weight.
- Read-only on day one: least-privilege read access per source, nothing changed in your systems.
Built from what you already run
Identity
- Microsoft Entra ID
- Okta
- 1Password
- Duo
Mail and collaboration
- Microsoft 365
- Microsoft Teams
- Google Workspace
- Slack
- Proofpoint
- Mimecast
Endpoint
- Microsoft Defender
- Microsoft Intune
- CrowdStrike
- SentinelOne
Cloud
- Microsoft Azure
- AWS
- Google Cloud
Network
- Fortinet
- Ubiquiti
- Cloudflare
- Palo Alto Networks
- Cisco Meraki
- Zscaler
Code and delivery
- GitHub
- GitLab
- Bitbucket
Service desk
- Jira Service Management
- ServiceNow
- PagerDuty
- Opsgenie
Logs and SIEM
- Datadog
- Splunk
- Elastic
Connectors not yet running in production are delivered within days of a pilot's request.
On another stack? Send it to us and we will say what works for it today.
Partners. Your analysts read the verdict and recommend in the same chat room as the customer; the customer keeps the approvals, the rollbacks and the record. Talk to us about a partner pilot.
Spot weak points
Accounts without a second sign-in step, outdated or unmanaged devices, risky sharing: found before someone else finds them.
A plain list, worst first, with what to fix and who owns it; the fixed ones drop off, the new ones are flagged.
- Accounts without multi-factor sign-in, dormant admin accounts, forwarding rules, access keys older than their rotation period.
- Devices unmanaged, out of date or without disk encryption; unknown devices signing in.
- Public sharing, open cloud storage and security-group openings, read from the same sources.
Catch attacks
Watched around the clock. Every event is looked at, not a sample, and the ones that matter are investigated in minutes.
Agents answer the events and incidents; when one needs a person, that person is asked, in plain words.
- Signals from identity, mail, endpoint, cloud and network are grouped into cases; an agent investigates each one read-only, with the evidence and the verdict written down.
- The person involved is asked "was this you?" in a private chat room; if nobody answers, the case keeps collecting and is escalated on a clock.
- No sample, no queue: every type of event and incident, the moment it arrives.
Every event is looked at, not a sample, and the ones that matter are investigated in minutes.
Every event and incident gets a look
Not a sample, not a queue, not only alerts: every type of event and incident is investigated the moment it arrives; in the example, from 02:14 to 02:26. You get a verdict in plain words and the reasons behind it. No time-to-verdict figure is published; a pilot measures its own.
A new device signed in from an unseen country and created a forwarding rule. Treat as account takeover; contain and confirm with the owner.
- New device, a country not seen before
- Sign-in method added four minutes later
- Forward-all inbox rule eleven minutes later
The person is asked, in chat. And if nobody answers, the case does not wait.
A sign-in from a new country on a new device at 02:14, then a forward-all inbox rule.
Was this you?
No answer.
- 02:27The room asks the owner: was this you? Answer: no.
- 02:31Within the identity response you enabled: sessions signed out, the inbox rule switched off, both announced in the room.
- 02:31A password change is proposed. It is permanent, so it waits for one click.
- 08:10A platform admin approves. The first plan expired after four hours; the renewed one carries fresh checks.
- 02:27The room asks. No answer. The case stays open and keeps collecting what it sees; the evidence is already on the case.
- 02:27Without response enabled, nothing acts. A proposed action is held for a person.
- 03:12The response clock started at hand-over. After 45 minutes with nobody accepting, the case is offered up your escalation ladder, with a chat line and a bell; the Escalate button says who will be paged.
- 06:27A held plan expires unapproved after four hours. A hold comes back to a person after 24 hours and is never lifted on a timer.
Act and report
Attacks are contained within the authority you grant, a person decides what matters, and you get a plain monthly report of what was protected and what to fix.
Nothing permanent happens without one click from a named person on your side; every action states how it is undone.
- Four response areas (accounts, devices, cloud, network), switched on one at a time, approval first; after 30 clean days an area may act on its own, with an Undo and an off switch.
- Protected accounts and servers are never acted on, even with approval; a plan not approved within four hours expires.
- A sealed record of who did what and when, a morning summary per team, a monthly report in plain words.
- 1
Responds within your rules. You hold the switches.
AI proposes the response. Independent controls enforce which actions are allowed, against which targets, and under which approvals. You enable response one area at a time, approval first; turning anything off asks for a reason and is written down.
What each action reverses or recovers, and what always waits for your click: on the security page.
Your switchesLook at every alertonAct on accounts, within the allowed listonAct on devicesask me first - 2
A record you can show
Who did what, when, in order. What prevents edits: the database refuses them for every role and logs every attempt. What detects tampering: every entry is sealed to the one before. What preserves an independent copy: a write-once copy at two providers for five years.
Case · historysealed- 02:14the deskNoticed the sign-in, opened the case sealed
- 02:30account ownerWas this you? No sealed
- 02:31the deskSigned out everywhere; recoverable by a fresh sign-in sealed
- 06:31the deskPassword-change plan renewed; the first expired unapproved after four hours sealed
- 08:10platform adminApproved the password change sealed
- 3
Every morning the short version; every month the plain report
One message per team each morning: what happened, what was done, what is waiting for you. Each month, a report in plain words: what was protected, what was fixed, what to fix next. Serious cases also open a ticket in IT Support, our own service desk, optional and bundled at half price.
case-0214-sign-inNightWatchApp8:00 AMOvernight, in short
- 1 account takeover: sessions signed out at 02:31; owner confirmed it was not them
- 1 decision waiting for you: a password change, plan renewed 06:31
- 2 more looked at and closed, reasons on the case
Account owner8:04 AMWhich inbox rule was changed, and is anything still being forwarded?
NightWatchApp8:05 AM- New device, a country not seen before
- Sign-in method added four minutes later
- Forward-all inbox rule eleven minutes later
Inbox rule switched off · 02:31 sealed1
