Pawtograder vs UniRubric

Pawtograder combines autograding, manual grading, office hours, discussion, a gradebook, and analytics for computer science courses. Each commit can trigger tests through GitHub Actions, with results shown by test in the grading interface. Rubrics support additive, deductive, and multi-level criteria, line-level annotations, and separate scoring for group members. The gradebook includes MathJS expression columns, drop-lowest calculations, CSV import, and a student what-if mode. Office hours queues support text, video, and screen sharing, and can show linked submissions and failing tests. Discussions can surface relevant threads for known errors and allow public or staff-only posts. Calendar feeds and Discord channel mirroring cover course events; LTI 1.3 integrations with Canvas, Blackboard, Brightspace, and Moodle are listed as roadmap items. Optional AI-generated test hints and an MCP server for compatible AI clients are also listed. Instructors can use the project-hosted community instance or self-host on Kubernetes with Helm charts. The project is GPLv3 open source. The free community instance is for evaluation; ongoing use at scale requires contacting the project.

UniRubric is a higher-education AI assessment platform that drafts rubric-anchored grades with quoted evidence for lecturer review. Lecturers can upload rubrics as DOCX, PDF, or plain text, then add student submissions individually or as a class. Each grade remains a draft until a lecturer reviews, edits, and approves it; it is not released to a student beforehand. Institutional integrations include Canvas, Moodle, and Blackboard through LTI 1.3, while the contracting page also lists Brightspace. A faculty lead dashboard shows rubric calibration across markers and can draft a meeting agenda and schedule a calibration meeting. Customer data, submissions, rubrics, and grading runs are region-pinned to Sydney, Australia, under standard operating conditions. The maker states that data is encrypted in transit and at rest, and customer work is not used to train or fine-tune models. Deleted assignments are removed from production within 24 hours and backups within 35 days; grading-run audit logs are retained for seven years. The Lecturer Free plan provides one lifetime marking job; Lecturer Premium costs $15 per month.