integrity
Catch every suspicious moment, in order.
Every paste, tab switch, idle gap, and AI corruption shows up as a marker on the replay scrubber — chronologically, never aggregated away.
4% first edit
11% AI corrupted
15% pushed back
23% paste 142ch
34% tab switch
41% idle 92s
58% 3/3 ✓
67% submitted
editpaste / AItab switchcaught / done
dashboard
Mission-control dashboard.
One screen for everything: live sessions, queued interviews, completed scorecards, with token spend and caught-rate trends across your team.
Search infra — onsite #2live—
Stripe payment flowpending—
Binary search — buggyended8.6
Aggregate orders queryended7.4
share
One link. No webcam. No install.
Send candidates a single URL — they create a free account in 10 seconds, then land in a real Monaco editor. No browser plugins, no screen-recording permission, no microphone. Just the code, the chat, and the moments that matter.
link is single-use · expires when session ends schedule
Schedule ahead. Or start cold.
Create a session and run it now, or assign a time slot and the invite link goes live automatically when the window opens.
JUN 2026S M T W T F S
123456789101112131415161718192021
guardrails
Stack guardrails. Or write your own.
Pick from five built-in policies — hints only, no full solutions, explain don't write, syntax only, open — and stack as many as you want. Custom presets and free-form rules let you shape the AI's behavior to match how your team actually interviews.
hints_only
no_full_solutions
syntax_only
+ custom rule
cost
Always free. Bring your own Anthropic key.
Acuity charges nothing. You pay Anthropic directly for the Claude calls your interviews use — a Haiku-powered 50-minute session lands well under a dollar.
billed by
Anthropic, not us
Typical 50-minute interview · claude-haiku-4-5 · ~5k tokens. Your key, your bill.
grading
Claude grades the judgement, not the typing.
Four dimensions: prompt quality, errors caught, code correctness, and how much they leaned on the AI. With a written justification per dimension.
7.6/10 · Alex ChenStrong hire
Prompt quality82
Caught AI errors75
Code correctness60
Approach88