Context

Context is the set of facts available to a run. It is assembled per run, from the ontology, the connected sources, and the boundaries you set.

Assembled, not dumped

A run does not receive your entire workspace. It receives what its scope allows and what the question is about. That keeps answers grounded and costs predictable.

Scope

Scope is set on the agent and narrowed per run. An agent scoped to invoices cannot read tickets, whatever the question asks.

Freshness

Context reads the most recent synced state. A source that has not synced since this morning answers with this morning's data, which is why the trace records the sync time it read.

Why an answer is wrong

Almost always because the fact was not in context: not synced, out of scope, or not modelled. Check those three before changing the brief.