AI Grader Pro vs Pawtograder

AI Grader Pro provides AI-assisted grading workflows for mathematics and other STEM subjects. Educators can create, refine, and apply rubrics, then review step-wise scores that account for partial credit and student solutions. The software processes question sheets and student answers in batches, and can export reports, summaries, and class-level outputs. Its local-first design is described as minimizing data leaving the user’s environment, but it can also send data to third-party AI services chosen by the user; those providers’ own terms and privacy policies govern processing. The product warns that AI-generated grading, feedback, and analytics may contain errors, leaving educators responsible for checking outputs and decisions. It runs on macOS and Windows. Recommended requirements are macOS 13 or later, or Windows 10/11 x64, with at least 8 GB RAM and 1 GB disk space. Full LaTeX output requires a LaTeX distribution. The Personal plan costs $120 a year and includes email support. The Team plan costs $299 a year, lists multi-user deployment and priority support, and is marked available in Q3 2026.

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.