Turning educational frameworks into usable data

The initial problem

The useful information already existed, but its format strongly limited how it could be used.

To find a specific item, a parent or teacher had to:

• identify the right framework;
• open one or more documents;
• find the relevant subject and level;
• browse pages until the target competence appeared;
• interpret the relationships between expected outcomes;
• copy the useful information before preparing an activity.

This search becomes especially heavy when building a learning path, checking that an activity is aligned, or regularly moving between several subjects and levels.

General-purpose AI tools raise another issue: without direct access to a structured source, they answer from general knowledge or from excerpts the user pastes in manually. The parent or teacher then has to copy framework content, explain the context, and check that the answer still matches the official expectations.

From document to usable data

MonPIF makes framework content available in a structured form.

Information can be organised and retrieved along dimensions useful for teaching work, including:

• learning level;
• subject;
• the competence or expected outcome being sought;
• relationships between the different elements of the framework.

That structuring is the real digitisation step. It does not change the official educational content. It changes how that content can be searched, linked, and used.

The same base can then feed several uses without asking every user to restart the extraction work.

A public Explorer for parents and teachers

MonPIF offers an online Explorer that makes browsing the frameworks simpler.

Users can target a subject, level, or competence without already knowing the exact document and page where the information sits.

This interface meets a concrete first need: making the frameworks more accessible to people who use them to prepare learning, build an activity, or check a pedagogical objective.

The Explorer is public, live in production, and used by both home-educating parents and teachers.

From search to creating learning activities

Structured data makes it possible to go further than consultation.

From a selected level, subject, and expected outcomes, MonPIF makes it easier to create learning activities aligned with the frameworks. Users no longer need to manually copy the relevant passages before every request.

The workflow becomes more direct:

• identify the level and subject;
• find the relevant competence or expected outcome;
• use that context to prepare a suitable activity;
• keep a clear link back to the original educational source.

Automation does not replace the judgement of the parent or teacher. It prepares the context and removes the repetitive work of searching, copying, and rephrasing.

Connecting ChatGPT to the frameworks through MCP

MonPIF also makes its data available to ChatGPT through a working MCP connector.

This connection lets the assistant query MonPIF when a user asks, for example, for an activity matching a particular level, subject, or competence.

Without that connection, the user must supply the needed excerpts themselves or accept an answer based on general knowledge. With MonPIF, the system can search the structured base for the relevant context before producing its response.

MonPIF thus becomes a specialised knowledge layer between the official frameworks and AI tools.

One infrastructure for several workflows

The project brings together four complementary uses:

• searching official frameworks;
• exploring content by subject, level, and competence;
• preparing aligned learning activities;
• making the data accessible from ChatGPT through MCP.

These uses rest on the same foundational work: turning content designed for reading into structured information that other systems can query and reuse.

A solution available in production

MonPIF is not limited to a technical demonstration.

• the platform is live in production;
• the Explorer is publicly available;
• home-educating parents use it;
• teachers use it;
• the MCP connection with ChatGPT is functional.

Together, these points validate both the usefulness of the interface for people and the ability to exploit the same data in automated workflows.

Impact

MonPIF brings several qualitative shifts:

• more direct search in frameworks that were previously hard to browse;
• exploration suited to level, subject, and competence;
• less copy-paste before preparing an activity;
• pedagogical context that is easier to reuse;
• a shared base for the Explorer, activity creation, and AI tools;
• the ability for other systems to use information that was previously locked in documents.

What this project demonstrates

MonPIF shows that useful automation sometimes starts well before the visible workflow.

As long as information stays locked in documents, people and software have to search, interpret, and re-enter it for every use. By structuring it once, it becomes possible to build several interfaces and automations on top of the same source.

The project demonstrates our ability to:

• turn complex documents into usable data;
• design an exploration engine for a non-technical audience;
• connect a specialised knowledge base to ChatGPT;
• build an application used in production by several profiles;
• keep the official source at the centre of the automation journey.