UDL 3.0 Model Context Protocol UDL-Con 2026
An open server that brings Universal Design for Learning into the AI tools you already use. Point it at the lessons and materials you have. It handles the framework, so you never have to become a specialist or learn to prompt.
Built on the Model Context Protocol (Anthropic, 2024). Works with any compatible AI client.
The paradox
No one asks a teacher to learn HTML to open a web page, or TCP/IP to send an email. The complexity stays hidden. Generative AI hides nothing: what you get depends on how well you prompt, which is really a programming skill. So the tools meant to make teaching inclusive reach only the people who can engineer a prompt.
The core idea
The server reads the lessons, slides, notes, audio and video you already use, maps them, and helps you redesign them against UDL 3.0. Then it turns your marking into feedback students can act on. Nothing new to author. Nothing new to master.
A plant's leaves capture sunlight and use it to turn water and carbon dioxide into glucose, the sugar that fuels growth. Same idea, read at your own pace.
The same explanation, spoken with captions, for learners who take it in by ear.
The same relationship, drawn, for learners who reason in pictures.
The workflow
Each module does one job. The output of the first feeds the next two.
Your notes, slides, manuals, audio and video become one queryable knowledge base, with prerequisites, objectives and UDL checkpoints made explicit. What was implicit becomes navigable.
It flags the UDL checkpoints you've covered and the ones you've missed, checks accessibility (WCAG 2.1: contrast, readability, alt-text), and suggests concrete alternatives for different learners.
It implements UDL Guideline 8.5, feedback oriented toward action, grounded in Bandura, Schunk, Hattie and Dweck. It recognizes effort, names a strategy, gives a next step, invites reflection.
→ The ontology from Module 1 feeds the other two. The effect compounds over a school year.
The feedback, made concrete
“6/10. You're not really a maths person, but the essay was okay. Study more.”
“You focused your argument well, and that structure lifted your score by 20%. Next step: rewrite one paragraph using the evidence-claim template, by Friday. What strategy will you reuse next time?”
The inclusion motif
Universal Design for Learning designs for every learner from the start: multiple means, one shared goal.
The WHY of learning
Sustaining motivation, effort and self-direction: the reasons a learner keeps going.
The WHAT of learning
Presenting information in more than one form (text, audio, diagram) so it reaches everyone.
The HOW of learning
Letting learners show what they know in more than one way: many routes to the same goal.
31 checkpoints, embedded across the three principles.
Control & trust
Control sits with the institution, not the model.
Consistent across every user. Quality no longer depends on prompt skill.
Inspectable workflows, not a black box.
Accept, modify, or ignore. The system analyzes, the teacher decides.
The team
A team from Pegaso University.
Development & AI Architect
Builds the server and the AI infrastructure.
umberto.barbieri@unipegaso.itUDL methodology
Grounds the system in Universal Design for Learning.
liadaniela.sasanelli@unipegaso.itScientific supervision
Leads the research direction and classroom validation.
raffaele.difuccio@unipegaso.itStatus
You'll be able to download it right here, very soon.
Coming soon