An open-source multi-agent framework for education. Specialized Claude agents coordinate across curriculum design, assessment, and learning paths. Implementable from your own GitHub repo.
Clone the EduForge repo from GitHub. Each agent is a self-contained Claude skill with its own SKILL.md, hooks, and tool configuration.
Define which educational agents you need. Set their roles, permissions, and how they communicate. The CLAUDE.md orchestrator ties them together.
Agents pass work between each other through structured handoffs. A curriculum agent designs, an evaluator reviews, an assessor builds tests. All coordinated.
Structures learning modules, sequences content logically, maps prerequisite knowledge, and builds skill trees for progressive mastery.
.claude/skills/curriculum-architect/Reviews educational materials against pedagogical standards. Checks clarity, depth, accuracy, and alignment with learning objectives.
.claude/skills/content-evaluator/Creates formative and summative assessments. Generates rubrics, quiz questions, and performance tasks aligned to curriculum goals.
.claude/skills/assessment-builder/Personalizes routes through the curriculum based on learner profiles. Adapts pacing, suggests remediation, and identifies acceleration opportunities.
.claude/skills/path-optimizer/Everything lives in your GitHub repo. Fork it, modify it, own it. No vendor lock-in, no monthly SaaS fees for the framework itself.
Use all four agents or just one. Each skill is independent. Compose the system that fits your educational context.
Agents propose, educators decide. The system augments expertise rather than replacing it. Every workflow has review checkpoints.
Not "AI for education" as an afterthought. Every agent is grounded in instructional design methodology: ADDIE, backward design, Bloom's taxonomy.
EduForge is building the agent infrastructure that treats educational design as what it is: a complex, multi-stakeholder process that benefits from coordinated intelligence.