Short version: Scolavo now has an MCP server — a first-party, licensed way to connect our entire curriculum to the AI tools your organization already uses. If your school, district, university, or company runs a chatbot or an internal assistant, you can now ground it in 3,680 class-lessons across 125 subjects, from Primary through College plus standalone Topics, with quizzes, transcripts, and structured lesson text — and a real license agreement behind every word of it.
The problem: your chatbot is only as good as what it reads
Most organizations that deploy an AI assistant discover the same thing within a week: a large language model on its own will answer confidently whether or not it actually knows. The fix is well understood — you ground the assistant in trusted content, so it answers from real material instead of improvising. But that raises a harder question: grounded in what? Most education content on the internet is scraped, unlicensed, or unstructured. Building an internal knowledge base on top of it means building on content you don’t have clear rights to — which is exactly the kind of thing that stalls a procurement review, and should.
What MCP is, in plain terms
MCP — the Model Context Protocol — is an open standard that lets AI assistants connect to external tools and content. Think of it the way you think of a web browser and a website: any browser can open any site, because they agree on a common protocol. MCP does the same for AI. A single MCP server can be used from Claude, from ChatGPT, from coding tools like Cursor, or from a chatbot your own engineers build — without a separate integration for each one. It’s the reason we could build one connection and have it work across the tools your staff already use.
What the Scolavo MCP server is
It’s a direct, first-party connection to the Scolavo curriculum. Not a scrape, not a mirror, not a third-party dataset of unclear provenance — the actual library, served by us: 3,680 class-lessons across 125 subjects, spanning five levels from Primary through College, plus our standalone Topics. Every class comes with its structured lesson text, its quiz bank, and its narration transcript. It is, as far as we know, one of the first full K-through-College curricula offered to institutions this way.
How it works for your team
For most staff, connecting takes about a minute: paste a URL into Claude or ChatGPT, sign in with the credentials we issue your organization, and you’re done. If your engineers run your own chatbot or internal assistant, they connect it the same way any MCP client connects — standard OAuth, nothing exotic. From then on, your AI can search the curriculum, pull a specific lesson, quiz bank, or transcript, and cite the source page on scolavo.com — so a teacher, student-services rep, or employee reading the answer can click through and see the original lesson.
The headline feature: whole-subject export
Searching and pulling individual lessons is useful for day-to-day work. But the feature institutions have asked us for most is bulk access — and the server delivers it. One call exports an entire subject: every class, every quiz, every transcript, and a manifest of the associated media, packaged as an organized bundle. Your team can index that bundle into a private retrieval system — the “RAG” setup most institutional chatbots are built on — and your assistant answers from licensed curriculum instead of whatever the open internet happened to contain.
What “licensed” actually means
This is the part we’d encourage you to read closely, because it’s where this differs from scraping a website or buying a dataset of unknown origin. Every organization signs a real license agreement. It permits internal use — grounding your assistants, building private indexes, serving your own staff and learners. It does not permit training AI models on the content or redistributing it. Your credentials are issued at the organization level, and every export is stamped with your organization’s license, so there is never ambiguity about what you hold or under what terms. When your legal team asks “where did this content come from and are we allowed to use it this way?”, the answer is one document long.
Control and transparency
Access is granted per organization, and only for what you actually license. If you only need one tier — say, High School — or even a single Topic, that’s all your credentials unlock, and that’s all you pay for. Your admins can see usage and quota standing directly through the connector itself, so there’s no separate dashboard to learn and no mystery about how much of your allocation has been used. And because access is managed on our side in real time, grants can be extended or revoked immediately — useful when staff change or an evaluation ends.
Privacy: content flows out, not in
The connector is a one-way street in the direction that matters. Curriculum flows out from us to you; nothing about your learners flows back. We don’t ingest student data through the MCP server, and it isn’t a student-facing product at all — it’s for your staff and your systems. Students never need a Scolavo account for your grounded chatbot to work. Whatever your assistant does with our content happens inside your walls, under your existing privacy posture, not ours.
Who it’s for
- School districts grounding a tutoring or homework-help assistant in real curriculum instead of the open web.
- Universities giving departmental chatbots reliable foundational material to cite.
- Enterprise learning and development teams who want an internal assistant that can teach — onboarding, upskilling, refreshers — from structured lessons.
- Teams already running a private RAG pipeline who need rights-clean educational content to index.
How to start
We offer evaluation licenses scoped to a single subject, so you can connect the server, run your real questions against real content, and show your stakeholders what a grounded assistant looks like before committing to anything. Pricing is annual and institutional — the kind that fits a purchase order — not per-call metering that surprises you at the end of the month. To begin, contact us through the support page and tell us a little about what you’re building; we’ll take it from there.
Questions we expect
- Which tools does it work with? Claude, ChatGPT (developer mode), Cursor, and any MCP-compatible host — one server, no per-tool integration.
- What does setup require? Standard OAuth with credentials we issue your organization. If your team has connected any MCP server before, this one works the same way.
- Do our students need Scolavo accounts? No. The connector is for staff and systems; no student accounts are involved anywhere.
- Can we train a model on the content? No — the license covers internal use and private indexes, not model training or redistribution.
- Can we start small? Yes. A single-subject evaluation license exists for exactly this reason.
We built Scolavo to be a school for the curious. The MCP server is that same library, offered to the institutions teaching curious people of their own — with the structure your systems need and the license your lawyers will ask about. If that’s you, we’d like to hear what you’re building.
