Search that finds it — and only shows you what you're allowed to see.
Search is the most persistent complaint about every incumbent chat tool. Aanty treats it as core infrastructure: hybrid keyword and semantic search, faceted by type, entity, and person, sub-second at scale — and trimmed to your permissions before results are ever ranked.
Operators when you know what you're looking for. Semantics when you don't.
Aanty combines keyword operators, semantic search, and facets in one query surface — so a precise search and a vague one both land.
Keyword operators
from:, in:, has:, before:, after: — the precision searchers already know how to use.
Semantic search
Find the right thread even when you don't remember the exact words that were used.
Faceted by type, entity, person
Narrow instantly by message type, linked entity, or who was involved — no query syntax required.
Sub-second at scale
Search is over-invested on purpose — it's the thing every incumbent gets complained about most.
The security promise: a query can never return what you can't see.
Trimming happens before ranking, not after — there's no window where a result briefly exists outside your access. Files, canvases, and transcripts live in the same index as messages, under the same rule.
One index, every artifact
Messages, files, canvases, and call transcripts are searchable together — not four separate silos.
Trimmed before ranking
Access is applied at query time, not as a post-filter — nothing you can't see is ever scored or shown.
Denied results stay locked
A result you can't open renders a locked empty state, never a partial title or snippet.
Defense in depth
The UI never grants what the database's row-level security denies — search enforces what the kernel already enforces.
Search reaches across every channel, entity, and artifact you can see.
Messages, files, canvases and transcripts are one connected, governed graph — so a single query traverses the relationships between them, not four disconnected search boxes.
Ask a question in plain language; get an answer you can check.
"What did we decide about the pricing page?" returns a direct answer synthesized from your graph — every claim traceable back to the message or document it came from.
- Answers, not just a results list. The graph is queried and summarized, not just matched.
- Every claim is cited. One tap opens the source thread or document behind any part of the answer.
- Permission-scoped like everything else. The answer is composed only from what you're allowed to read.
"What did we decide about the pricing page?"
We're keeping unlimited history on every tier, including free — logged as a decision on 2 Jun, owner @jitesh.
A cited answer composed only from what you're allowed to read.
This is what a question returns: a synthesized answer, source pills you can open, and nothing pulled from a channel your permissions don't reach.
The Q3 security audit is on track. 12 of 15 findings are remediated1; the remaining three are scoped to the R-142 vendor-access work due Jul 182. Legal signed off on the retention change on Tuesday3.
Search that remembers what you care about — and builds a memory of its own.
Saved searches & subscriptions
Save a query, or subscribe to one — "alert me when anyone mentions churn and enterprise" — and let it watch instead of you.
Org memory
Entity profiles for projects, customers, and systems are auto-built from conversation — visible and editable, not a black box.
Ask your own question about your own graph.
Bring a question your team keeps re-asking. We'll show it answered, cited, and trimmed to exactly what you're allowed to see.