Asked, answered.

Assessly is an evaluation platform for programming education. Every submission gets two scores: whether the code is correct, through sandboxed test execution, and whether the student understood it, through an AI-graded explanation of their approach. Universities use it to run exams whose grades certify understanding, not just output.

Universities and colleges first, departments grading programming courses and placement cells preparing students for hiring. Engineering teams also use it for technical screening. Students inside those institutions get practice problems, structured lessons, AI mock interviews, and progress tracking.

Six products on one platform: Assess for proctored exams with dual scoring, Explain for AI-scored walkthroughs of a student's reasoning, Interview for AI mock interviews with voice, Learn for daily practice that builds the habit, Grow for placement-readiness analytics per student, and Speak for listening, group discussion, and professional writing.

Five (Python, JavaScript, Java, C++, and C). Every run executes inside an isolated sandbox on dedicated execution infrastructure, kept separate from the application and your data.

Assessly is invite-only. Students and staff join through an invitation from their institution or a class enrollment code. If you're evaluating Assessly for an institution or a team, book a demo and we'll set up a pilot.

Every submission produces two independent scores. The code score comes from test execution, public and hidden cases, run in a sandbox. The explanation score comes from AI grading the student's own walkthrough of their approach. A correct solution with a hollow explanation is exactly the gap dual scoring surfaces.

Students explain their approach in their own words: why this data structure, what the complexity is, which trade-offs they considered. The AI grades that reasoning for clarity and depth against the instructor's rubric, independently of whether the tests passed, and returns written feedback alongside the number.

Rather than asking you to trust a single number, every AI score ships with the written reasoning behind it, so an instructor can check the judgment directly. The grader is calibrated against human-graded reference examples, and instructors can override any AI score manually. Overrides are recorded alongside the original.

Yes. Each assessment sets its own split. 70% code and 30% explanation is a common starting point, but instructors choose the weights and the final score respects them.

Reattempts are off unless the instructor allows them on a given assessment. When they're on, the instructor also picks the scoring policy: highest score, latest attempt, or the average of all attempts.

A preflight check. Students see the assessment overview, pass system checks, test the camera and microphone when proctoring requires them, and give explicit consent to the monitoring that's enabled. Only then does the session begin, in fullscreen.

The work is safe. Code and answers autosave as drafts, and the session freezes on a clear overlay until the connection returns. Disconnects are logged as events, not punished. The timer runs on the server, so if an outage costs real time, the instructor can extend the deadline while the attempt is still live.

Visible test cases are shown while solving. Hidden test cases never reach the browser. They're applied server-side at submission, so a solution that hardcodes the visible cases doesn't pass.

Yes, in three ways: instructors set the time limit per assessment, can allow a late-join grace window, and can extend a deadline mid-attempt. The extension reaches the student's session automatically, without restarting anything.

Yes. Learn offers structured lessons and a practice library with the same editor, the same sandboxed execution, and the same AI feedback as a real assessment, so on exam day, the interface is never the surprise.

A catalog of modules instructors enable per assessment: fullscreen enforcement, tab-switch tracking, copy-paste blocking, devtools detection, single-tab and single-device locks, browser fingerprinting, typing analysis, network anomaly detection, webcam proctoring with face detection, ambient audio monitoring, and event-triggered video clips. The configuration is sealed once the attempt starts.

As little as each module needs. Tab switches and fullscreen exits are stored as counts, not screenshots. Paste events keep a length and a SHA-256 hash, never the raw text. Typing analysis keeps timing summaries and discards raw keystrokes. Webcam proctoring captures frames only when an anomaly is detected, and audio monitoring stores a speech-likelihood score. No audio is ever recorded.

No. The philosophy is flag, not block. Violations are logged with context (what happened, when, and the evidence) and surfaced on the instructor's dashboard in a form a misconduct board can actually use. The decision always stays with your faculty.

Detection runs inside the evaluation pipeline. A flagged explanation carries a probability and the specific indicators that triggered it, and is queued for instructor review. Like every integrity signal, it informs a human decision rather than making one.

TLS 1.3 in transit, AES-256 at rest, and row-level security in the database, so each institution is an isolated tenant. Code execution runs on separate, dedicated infrastructure. The full security posture and subprocessor list are public on the security page.

Yes. Account deletion is self-serve in settings and requires the account password plus a typed confirmation. Retention of institutional records, submissions and grades, is governed by the institution's agreement, since those are the institution's academic records.

By the shape of your program: the size of the cohort, the products you enable, and how much of the integrity stack you run. We scope it on a short call and put the quote in writing. We don't publish prices, because configurations vary too much for a fixed public plan to be honest.

Yes, most institutions start there. A guided pilot runs on a real cohort for a term, so your instructors see dual scoring, proctoring, and the dashboards on their own students before any annual agreement.

Three routes, usually combined: token-based email invitations, class enrollment codes for self-enrollment, and CSV bulk import with student-ID validation against your institution's format. Faculty and instructor seats are included on every plan.

SSO and user provisioning are part of the Enterprise tier, and LMS integration is on the roadmap. If integration is a procurement requirement, raise it on the scoping call and we'll walk through your environment together.

Code execution is unlimited under fair use, and running code never consumes AI allowance. AI grading is included generously; only AI mock interviews are metered, and they can be expanded anytime. Limits are soft, so students are never blocked mid-assessment.