Problem

Today, to answer a question like "what do we have on this topic?", an agent has to query the API of each website. While this works for a small number of projects, it does not scale as a primary method for interacting with the cluster.

We need a single information entry point that first allows the agent to understand where the relevant knowledge is located, and then, if necessary, proceed to the specific source.

Working Hypothesis

A possible solution is a central hub that aggregates local knowledge maps of websites and source metadata. It does not necessarily have to store full copies of materials or become a shared CMS. An index/map model, where the full content remains on the original sites, looks more interesting.

This direction requires validation: what exactly should be centralized, what data is more efficient to keep local, how to preserve provenance, and how to avoid creating a second competing source of truth.

The central hub is part of the general circuit and depends on the understanding of the local knowledge map, the website registry, and synchronization.

Desired Result

The agent should be able to get an overview of relevant knowledge and its sources across the entire available circuit in a single request, without having to perform dozens of repetitive queries to individual websites beforehand.