Integrity, not invigilation.
The sector reached for more surveillance. We reached for better assessment.
This page is for the people in your institution who need the receipts. What Sentinel does, what it does not do, and the design choices behind both.
Integrity is built in two layers. The first does most of the work.
The assessment itself.
A Meandrix Encounter is dynamic by design. The learner’s decisions change the course of the encounter, shaping the situation, the information available, the risks that emerge, and the decisions that follow.
That is the first layer of assessment integrity. Meandrix does not rely only on restricting behaviour during an assessment. It strengthens integrity by changing the nature of the task itself: from recall to applied judgement in context.
The proportionate second layer.
Sentinel adds integrity friction when the stakes require it. It does not replace good assessment design. Instead, it strengthens the conditions around the assessment by making dishonest behaviour harder to hide and less attractive to attempt.
When enabled, Sentinel records key session conditions and creates a timestamped evidence trail. The experience remains low-friction for honest learners.
Opt in. Only when it matters.
Sentinel is not on by default. Educators choose when to enable it for assessments where an integrity record is warranted.
Formative practice runs without it. Asynchronous learning runs without it. The disclosure, fullscreen requirement, and session record only appear when the assessment requires additional integrity assurance.
The full list. In plain terms.
Disclosure and consent
Before any monitoring begins, the learner sees exactly what is tracked and must confirm they understand.
Single-display check
A second monitor is detected and the assessment cannot start until it is disconnected.
Fullscreen mode
The assessment runs in fullscreen. Exits are recorded and the learner is returned with a notice.
Tab and window focus
Leaving the assessment window is logged and the content is obscured until the learner returns.
Screenshot attempts
Screenshot attempts are detected via screen-capture key combinations and focus-loss patterns consistent with screen-capture tools. Each attempt is logged as a high-risk event and flagged for educator review. The device clipboard may be overwritten with an integrity notice during the assessment.
Developer tools
Attempts to open browser developer tools during the assessment are detected and logged.
Right-click and selection
The context menu and text selection are disabled during the assessment.
One device per attempt
Each attempt locks to a single browser session. Concurrent logins from another device are flagged.
Session timeout
Extended inactivity pauses the session. The learner acknowledges to continue, and the gap is recorded.
Identity visible throughout
The learner's name, institution, and assessment title remain on screen during the entire session.
Timestamped event record
Every event above is stored against the attempt and available to the educator and institution.
The line we will not cross. Equally explicit.
What we choose not to do matters as much as what we do. Sentinel is built so a learner doing honest work never has their privacy or dignity compromised.
No webcam access. Sentinel never sees the learner.
No microphone access. The learner is never recorded.
No facial recognition or biometric identification.
No keystroke logging beyond the specific integrity events listed above.
No AI scoring of behaviour or inferred intent.
No human proctor watching the learner in real time.
No location tracking. No GPS, no IP geolocation profiling.
No browser history, no social media, no third-party data collection.
No data shared with anyone outside the institution that owns the assessment.
The learner sees everything first. Then they decide.
Nothing is monitored covertly. Every Sentinel session begins with a full disclosure of what is tracked, why it is tracked, and what happens to the record afterwards.
The learner has to explicitly confirm they understand. The assessment cannot proceed without that consent.
This is not a polished compliance gesture. It is the central commitment. The work asked of the learner is honest work, and the conditions under which they do it are visible to them from the start.
Academic Integrity Monitoring
This is a monitored summative assessment
The following activity is tracked and reported to your educator:
- • Tab switches and window focus loss
- • Fullscreen exits
- • Screenshot attempts
- • Developer tools access
- • Clipboard may be overwritten with an integrity notice
Monitoring is a requirement of your institution's academic integrity policy. This assessment must be completed in fullscreen mode.
Where the record lives. And where it stops.
Stored against the attempt
Every Sentinel event is timestamped and attached to the assessment attempt record. There is no separate behavioural profile, and no learner is tracked across attempts.
Owned by the institution
The integrity record belongs to the institution that owns the assessment. The educator and authorised institutional admins can review it. No one outside the institution has access.
Never used to train AI
No learner data, including Sentinel records, is used to train any AI model. This is contractually enforced and mirrors the no-training policy of Anthropic, the AI infrastructure provider Meandrix uses for scenario generation.
Hosted in Sydney, Australia
Meandrix runs on Supabase infrastructure in the Sydney region. Learner data does not leave Australia.
Retention follows your records policy
Sentinel records are retained for as long as your institution's records management policy requires. Australian higher education providers operate under state-based records frameworks (such as NSW GA47 under the State Records Act 1998, or Victorian PROS 16/07 under the Public Records Act 1973), and the relevant minimum retention class is determined by the institution.
Auditable on demand
Any attempt record can be exported in full for an Academic Integrity Committee, an accreditor, or a student appeal. The record is the same one the educator sees.
Built to the standards your accreditors expect.
TEQSA 2023: Assessment reform for the age of artificial intelligence
Meandrix is built around the principle that assessment should evidence what the graduate can actually do, not just what they can recall. This aligns with the foundational guidance issued to all 203 Australian higher education providers.
TEQSA 2025: Enacting assessment reform in a time of artificial intelligence
The 2025 guidance endorses program-wide reform, unit-level assurance, and hybrid models. Meandrix supports each of these pathways with applied decision-making assessments that work alongside written and OSCE-style instruments.
ANMAC accreditation
The Australian Nursing and Midwifery Accreditation Council expects evidence of competency-based assessment. Sentinel-protected Encounters and Consults produce timestamped records of applied judgement that map directly to competency standards.
NMBA standards for practice
The Nursing and Midwifery Board of Australia's standards for practice cover clinical reasoning, decision-making under pressure, and professional accountability. Each is observable in the way a learner moves through a Meandrix Encounter or Consult.
Paramedicine Board accreditation
The Paramedicine Accreditation Committee expects assessment of decision-making in unsupervised, time-pressured environments. Consult-mode scenarios with critical-error termination map closely to this requirement.
Want to see how Sentinel works? Book thirty minutes.
I can walk you through Sentinel in detail, show how it works in practice, and answer the questions that matter in your context.
Book the conversation