by Shadecode




What it does
Cortex Study Room is a multi-agent learning system that turns a learner's goal into an adaptive study session. Four specialised agents work around the same learner state: Research maps the knowledge needed, Instruction builds the lesson, Review detects gaps and misconceptions, and Assessment tests understanding. Their findings are shared through an event-driven state layer so a detected gap can trigger another agent to adapt its work instead of following a fixed sequence.
How the agents run concurrently
When a study room starts, Researcher, Tutor, Challenger, and Examiner are launched concurrently rather than as a fixed pipeline. They share a typed runtime state containing the learner goal, concepts, lesson, detected gaps, questions, answers, mastery, and current plan. Agents react to events in that shared state: for example, Challenger can emit a concept-gap event while the Tutor is working, causing the Tutor to revise the lesson, while the Examiner uses the resulting state to create a diagnostic question. The repository also includes a Mozaik runtime adapter using @mozaik-ai/core for the concurrent/reactive agent architecture.
Newsletter
For developers who want to learn how to build self-organizing agents.
Join our online hackathon
Join our online hackathon for building multi-agent systems.