Omniscol's general philosophy

Before the features, a few principles structure the Omniscol experience and explain why the software makes certain decisions rather than others.

The business context

A timetable is not just a table to fill in. It is a constrained optimization problem:

• many constraints, often implicit;
• constraints that sometimes contradict each other;
• high operational stakes: the school year start, the semester or the period must work.

Omniscol is designed as a constraint-based planning assistant: it speeds up the work, flags impossibilities and leaves business decisions to the user.

What Omniscol is not

Omniscol is meant to interface with other systems via imports, exports, iCal, a documented REST API, OIDC / SSO depending on the contract, and synchronization with external systems. It does not try to replace the business tools that cover other scopes. See also integrations.partners.

The five founding principles

1. Data before computation

You cannot generate a reliable timetable without healthy data: sites, rooms, teachers, classes, groups, subjects, hour volumes, availability and constraints.

2. Separating structure from daily life

Building a timetable and living with a published timetable are two different activities. The first mostly happens in Timetable management, the second in Timetable, Absences and Dashboard.

3. Hard constraints and optimization constraints

Some constraints make a timetable invalid and are treated as strict constraints: a teacher in only one place at a time, a class without collisions except for groups in a division, a standard room assigned to a single lesson, unavailability, capacities, material resources and inter-site travel.

Other constraints are used to improve the solution: undesirable availability, gaps, day balance, pedagogical order between subjects, hour maxima or minima, number of days of presence. They are evaluated as penalties that the solver tries to reduce.

4. Drafts and versions

An unpublished timetable has no impact on what users see. You can duplicate, test, compare, lock some lessons, run another generation and publish only when the result is right.

Snapshots, when enabled, add a safety net for going back or restoring some data.

5. The user stays in control of the data

Omniscol provides full JSON export, business exports per screen and reversibility. Automations do not replace human decisions.

Conflicts are information, not errors

In manual entry, Omniscol flags conflicts but lets the user decide. This is useful for real special cases: a room for extra-time exam arrangements, a deliberate overrun of a theoretical capacity, assigning a room outside its specialization for a local reason.

In automatic generation, the solver does not deliberately produce major collisions. If no complete solution is found, it keeps the best computed timetable and leaves the lessons impossible to place in the list of unplaced sticky notes.

Locking a lesson anchors a chosen placement: the next generation adjusts the other lessons around that lock.

Multi-timetable as a modeling principle

Omniscol is natively multi-timetable. Several timetables can coexist for the same school:

• distinct school years;
• semesters or periods;
• logical separation by campus, department or cycle;
• a combination of different types, for example recurring weekly and dated calendar.

In Premium, several timetables can be active in parallel — a specific activation is also possible on some Standard accounts. A school can have a preparatory cycle with two weekly timetables, one per semester, a graduate cycle on a non-recurring calendar and simultaneously, an 18-month ExecMBA program that starts and ends offset from the other schedules and from the school year. With Omniscol, you can create each timetable separately while benefiting from cross-conflict management during planning, then activate everything in parallel, simultaneously, so that the final result is a merged view, transparent to end users.

Optimization AI and external agents

The solver is a neuro-symbolic Monte-Carlo metaheuristic optimization AI. The engine explores placements under constraints, looks for a valid solution then optimizes the penalties, with impressive speed (e.g. less than a minute for a middle school with 16 classes). If no valid solution is found, a relaxation system sacrifices lessons intelligently, and still proceeds to optimize the result. Omniscol is an AI-first company, born from a long "deeptech" R&D phase.

Omniscol also has other symbolic-AI algorithms, to check conflicts, detect configuration inconsistencies, make improvement suggestions or configuration alerts, pre-filter the best solution from a list of proposals, or automatically allocate a subset of rooms to a subset of lessons, all in the browser, immediately.

Omniscol also exposes API tools to compatible agents via MCP when the option and the rights are active. This makes it possible to interact with Omniscol in natural language, to extend what the graphical interface offers, by fetching the relevant data directly from the Omniscol server. These agents act within the scope of the provided token; read-only uses are the safest, and any write action must remain verified by the user. Currently, the best AI agent for interacting with Omniscol is Claude, in its Desktop version.