Tools

When you connect SubVault, your AI gets two verbs — vault for saving knowledge, remember for recalling it — plus a small third tool, corpus_stats, for counts. You don't call them directly — your AI does, based on what you ask.

vault

vault
Save knowledge from a conversation. Your AI extracts the structured items worth keeping — facts, decisions, action items, people — and saves them to your vault. Your full transcript is never sent — only what your AI explicitly vaults. Try: "Vault what we just decided about the schema migration." Try: "Remember that we use Postgres for our production database."

remember

remember
Pull relevant context. Your AI calls this when you ask about your past work, your projects, your preferences, or anything else that lives in your vault. It returns a curated bundle scored for the current conversation — not a full dump. Try: "What was I working on last Tuesday?" Try: "What's the status of Project Hyperion?" Try: "What do you know about my tech stack?"

corpus_stats

corpus_stats
Count what's inside. Returns totals — facts, decisions, action items, people, documents. Numbers only, never content. Try: "How big is my vault?"

That's it

Two verbs plus a stats tool. Your AI decides which to call based on what you ask. You don't pick.

Vault stats come from corpus_stats — just ask, e.g. "how big is my vault?"

How it flows

Most chats look like this:

  1. You ask a question that needs context ("what did we decide about the API rate limits?").
  2. Your AI calls remember with a situation string describing what you're working on.
  3. It gets back facts, decisions, and action items relevant to that situation.
  4. It uses that context to answer.

Saving works the same way: you say "vault this" or "remember that..." and your AI calls vault with the right content.

To keep a topic out of recall or have an old fact dropped, email info@subvault.ai.