A Disclosure at 9pm Should Reach a Person
Students ask for help in indirect ways, often on a school device, often at night. Keyword alerts put that signal in front of a named human within seconds — by email with full context, and by text message when it matters.
Why You Need This
A search for “I want to disappear” may be the only signal a student gives. The hard part was never detecting the word. It is making sure a person sees it, quickly, with enough context to know whether to act tonight or in the morning — and without so many false alarms that staff stop reading.
Detection happens on the device. The managed extension watches text typed into search boxes and form fields against a keyword list the district controls. Each keyword carries its own severity — low, medium, high or critical — which is what separates a health-class assignment from a statement of intent.
Every alert is routed, never dropped. A rule for that school, otherwise the district-wide rule, otherwise district administrators. An alert can never route to nobody.
The email carries the investigation, not just the word: the student, the building, the time, the full search text, and every other alert for that student in the last 30 days — or a plain statement that there are none and this is an isolated event. That is the difference between “a word was flagged” and “this is the third time in two weeks.”
The text message carries no student identity. Severity, building, time and a link, sent from a number verified with the mobile carriers so it is not silently filtered as spam. A text is not a secure channel, so identity stays behind the login.
- ✓ Severity you set, per wordEight categories including self-harm, violence, bullying, drugs and network bypass — each word placed where your district wants it, not where a vendor decided.
- ✓ Context in the alert itselfThe student's last 30 days of alerts arrive inside the email, so staff can judge a pattern without opening a console.
- ✓ Email and text messageText delivery on a dedicated, carrier-verified number. After-hours disclosures reach a phone, not an inbox nobody opens until Monday.
- ✓ Recipients per buildingA counsellor sees their own students and no one else's. Quiet hours are set per rule, with critical severities deliberately breaking through.
- ✓ Storm protectionOne incident cannot flood a phone. After the first alert in a building at a given severity, the next ten minutes are counted rather than sent.
- ✓ A daily digestThe previous 24 hours per building, for staff who want the pattern rather than each event.
Common Use Cases
The late-night disclosure
A student types something concerning at 11pm on a school Chromebook. The counsellor you named gets a text within seconds, and an email carrying the full context and the student's recent history. Nobody waits for Monday.
Telling research from risk
Health classes generate searches that look alarming in a keyword report. Severity levels and context scoring keep the health-class searches quiet so the real ones still get read.
Seeing a pattern form
One alert is an event. Three in two weeks is a student. Because every notification carries the last 30 days, the pattern is visible in the alert itself rather than assembled afterwards.
Frequently Asked Questions
What exactly is monitored?
Text typed into search boxes and form fields on the school-managed Chromebook, matched against the district's keyword list. Documents, private accounts and personal devices are not read.
Who decides which words matter, and how serious they are?
The district. Every keyword carries a category and a severity you set, and you can start from a recommended list and edit it. Districts commonly place clinical terms like "depression" lower on purpose, because they are the single biggest source of noise in a school setting.
Does a text message contain the student's name?
No, by design. The text gives severity, building, time and a link. A phone can be on speaker or read by someone else, so identity and the matched words stay behind the dashboard login. The email, which goes to the staff you name, carries the full detail.
Do you use AI, and does it decide what we see?
Every alert is additionally scored 0–100 for real-world urgency with a short written reason, weighing the matched term, the full search text and the time of day. That score is advisory: it annotates the alert for whoever reviews it and does not route, downgrade, delay or suppress anything. We are not willing to let a model decide whether an adult hears about a child.
What stops one incident from flooding everyone's phone?
A throttle. After the first delivery for a building at a given severity, further alerts in the next ten minutes are counted rather than sent, so a single event does not turn into thirty texts.
Can we set quiet hours?
Yes, per rule — and critical severities bypass them by default, because the reason to have alerting at all is the 2am case.
Make sure the alert reaches a person
Severity you control, routing per building, and delivery by email and text — with the context staff need to act.