Search
search
An agent reaches for this when it needs the right document fast — ask in plain words, get the best matches with a source and a match score for each.
Searching your documents…Five real runs against the documents you just indexed — nothing here is canned. Each card names the tool your agents call to do the same thing.
{# Each card owns a fixed-size result slot (kx-tour-result) that is present from first paint — results render INSIDE it, so the cards below never move when a run finishes (#488 layout invariant). #}search
An agent reaches for this when it needs the right document fast — ask in plain words, get the best matches with a source and a match score for each.
Searching your documents…prep
An agent calls this before starting a task — it gets back a short, compact context pack on the topic, built only from your documents. This is what your agent receives.
Building the context pack — takes up to half a minute. This is a real run against your documents.memory_write
When an agent learns something worth keeping, it saves a memory — the note lands in your knowledge store and search finds it seconds later.
Saving it, then searching for it…brief
An agent starts its session with this — a short summary of recent activity, so it picks up where the team left off instead of starting cold.
Writing the briefing — takes up to half a minute. This is a real run against your documents.timeline
When the question is about time — what happened last week, what changed in May — an agent uses this to get results that respect dates.
Looking through time…