Token-saving retrieval: ONE query returns a compact context pack of only the memories/files that matter, fitted to a token budget. Call this INSTEAD of re-reading files or scrolling history — it replaces the re-reading tax with a targeted answer. Top hits come back as capped previews (~1800 chars), the rest as gists — cairn_read returns any node's complete text (raise max_chars for very long nodes).

The creative complement to cairn_fetch. Walks the memory graph outward from the query's best hits, preferring WEAK ties that cross topic boundaries — surfaces adjacent ideas and unseen connections (a past project, an old conversation, an idea from a different domain) that similarity search will never return. Use when brainstorming, stuck, or asked for fresh angles. Returns gists — cairn_read pulls any hit in full (raise max_chars for very long nodes).

Ranked hybrid search across the whole vault. Returns ids + gists + scores (cheap) — an index, not the text. Pass the ids that matter to cairn_read for their full text — raise max_chars for very long nodes (cairn_fetch for a budgeted context pack).

Write a memory to the vault — ambient capture. Use for decisions, insights, ideas, open items, warnings, procedures discovered while working. The vault remembers across sessions and models. Write the complete salient fact, decision, warning, or open item WITHOUT truncating it — but do not paste transcripts (turns are captured separately). For a large artifact, save the file and note its path and purpose.

Session-start context: the most recent compiled PROTOCOL digest plus open items. Call once at the beginning of a session to inherit prior context.

The working set: recent decisions, open items, and warnings across the vault — what's live right now. Returns gists — pass ids to cairn_read for full text (raise max_chars for very long nodes).

Read specific nodes IN FULL by id — the follow-up to the gists that recent/search/fetch return; this is the verbatim layer (long turns carry their complete text here). Accepts full ids or unambiguous prefixes, up to 8 per call; max_chars dials the per-node body budget. Unlike fetch/search this needs no embeddings, so it reads nodes written moments ago.

The LIVE tail of the vault — newest nodes first, ALL kinds including conversation turns, embedded or not. This is how you see what just happened (fetch/search can't see un-embedded nodes until the nightly sleep). Filter by kind, session prefix, or a contains substring; follow up with cairn_read for full text (raise max_chars for very long nodes).