AI-evaluated student engagement

Quiet, careful attention for asynchronous classrooms.

ClassLens turns every activity into a watchful colleague. Students work; the platform notices when they’re focused, when they’re stuck, when they wander — and writes you a careful, minute-by-minute account.

Webcam stays on-deviceFree for classes under 30 studentsSOC 2 in progress
Replay · Priya Raman
Reverse a linked list, iteratively
minute 08 of 22
screen recording · 22:14
Latest minute
·Stuck·score 41
+08:00

Long pause on the same loop body. Mouse hovers but no edits.

Engagement timeline
Focused
Drifting
Stuck
Off task
Away
How it works

Three steps from objective to replay.

ClassLens sits between you and the work itself. You don’t grade what you didn’t see — you grade a careful account of what happened, written in language a teacher would actually use.

01
Generated checklist · 6 steps
S01Read the brief
S02Sketch the algorithm
S03Implement reverseList
S04Run the test suite

Write the objective.

Describe the work in plain language. ClassLens drafts a structured checklist — labels, completion criteria, and what evidence to look for — that you edit before publishing.

02
Invite · lnk-2026-bx9
Reverse a linked list, iteratively
screenwebcam

Share the invite link.

Students open the link, see what's about to happen, grant screen + webcam permission, and start. ClassLens captures screen in 60-second chunks; the webcam runs a face model on-device.

03
Minute 12 · Stuck → Focused

Spots the off-by-one — the next pointer is being read after curr is mutated. Fixes the order of assignments.

Review the replay.

Each attempt arrives as a colour-coded timeline, an AI-written narrative summary, and a clickable list of key moments — the off-by-one bug, the tab switch, the green test run. Async-first, not Zoom-with-extra-steps.

Methodology

How a minute of work becomes a sentence you can trust.

ClassLens is a four-stage perception pipeline. Each stage throws information away on purpose — so by the time a paragraph reaches you, the only thing left is what a careful colleague would have noticed if they were sitting next to the student.

00
Input
raw streams
capture
01
Observe
60s windows
extract
02
Infer behaviour
6 states
reduce
03
Narrate
1 paragraph
deliver
04
Output
for teacher
SCREENstudent's tabWEBCAMon-devicenever uploadedL1 · OBSERVEscr_chunk60scam_frames12 fpstab_focuseventsL2 · INFER BEHAVIOURfocuseddriftingstuckoff-taskawaysubmittedL3 · NARRATE“Spots the off-by-one. Fixes theorder of assign-ments…”TEACHERreplay + summaryDEVICE ← | → CLOUDstaysuploads
Each stage throws data away →~10 MB of pixels~7 numbers1 of 6 states1 paragraph
Privacy as posture

The webcam never leaves the device.

We run the face-landmark model in the student’s browser. The only things that travel to our servers are numerical features — attention percentage, face presence, head pose — and a single low-resolution keyframe per minute, kept only for training the perception layer.

No raw video. No audio. No keystroke logging. No DOM scraping. Screen capture is exactly what the student picks at the share-prompt — they choose the surface.

STUDENT'S DEVICE
STAYSWebcam streamraw video
STAYSFace modelMediaPipe · 468 landmarks
STAYSAudionever captured
features
only
CLASSLENS SERVERS
UPLOADpct_attentive0.92
UPLOADpct_face1.00
UPLOADhead_pose(yaw, pitch)
UPLOADscreen chunk60s WebM
UPLOADkeyframe1 / minute · 240×180
For teachers

Built for the teacher who reads every attempt.

“I used to spot-check submissions and hope. Now I read a one-paragraph summary per student and I know — to the minute — where each one got stuck. Office hours stopped being triage.”
Iris Hartmann · CS & Math, Parkhill Upper School
What you get
Behavioural, not affective.
Six observable states — focused, drifting, stuck, off-task, away, submitted. Never 'confused' or 'frustrated.'
Replay over real-time.
Async-first review is the default. Live dashboards exist; they're not the centerpiece.
Key moments, not heatmaps.
Each minute carries a few clickable moments — the bug found, the tab switched, the test run that turned green.
AI-written summaries.
One careful paragraph per attempt, in plain teacher-speak. Edit before sending or keep it private.
Privacy as posture.
Webcam never leaves the device. Audio is never captured. Students see exactly what's recorded.
Pricing

Free for small classes. Honest for big ones.

Classroom
Free
for classes under 30 students
Start free
  • Unlimited activities
  • Live + replay review
  • AI summaries
  • 30-day history
MOST POPULAR
Department
$8
per teacher / month
Start 30-day trial
  • Everything in Classroom
  • Unlimited students
  • 1-year history
  • Roster import (CSV, Clever, Google)
  • Shared activity library
School
Talk to us
annual, per school
Contact sales
  • Everything in Department
  • SSO (SAML, OAuth)
  • SOC 2 report
  • Audit log
  • Dedicated success manager

Try it with one activity.

Five minutes to publish, sixty seconds to see the first replay. Free for your first class — no card, no demo call.