Skip to main content

What your assistant can do: the JustOnce tools explained

The tools a connected assistant sees, in plain words — so you know what it's doing when it says 'checking your memory'.

A
Written by Aaron Wittman

Your assistant never sees your vault directly; it calls tools. Here is what each one does. What a connected assistant does is recorded in your Ledger.

Remembering and recalling

  • get_core_profile — loads your standing profile once at the start of a session: verified, high-confidence facts (nothing sensitive), the people and things that come up most, and your active projects.

  • search_memories — finds memories relevant to a question, each with a verdict on whether it may be stated as fact (see Confidence, conflicts and "assertable"). It also finds matching memories other people shared with you, and says who shared them.

  • get_context — a ready-made block of what your own vault holds on a broad topic ("what do you know about my car?").

  • add_memory — stores one fact. import_memories stores many at once (up to 50 per call).

  • get_memory — reads one memory in full: entities, action items, confidence, timestamps and any attached files.

  • update_memory — corrects a memory, keeping the old value in history. forget_memory erases one for good; there's no undo. Your assistant should confirm first, and erasing several in a row needs your explicit go-ahead.

Time and relationships

  • get_as_of — what was true about something on a given date.

  • get_memory_history — every value a memory has held and when JustOnce believed it.

  • get_relationships — walks the graph of how the people, organisations and places in your memories connect ("how do I know Priya?").

  • find_related_entities — lists the people, organisations, places and accounts your memories mention, or checks how well-supported one stored value is (an address or account number) before your assistant relies on it.

Documents

  • store_document — saves a small file the assistant holds, plus the facts it read from it. JustOnce doesn't OCR, so your assistant passes along what the document says.

  • request_upload_link / check_upload_link — gives you a one-time browser link to upload an original file (up to 25 MB; it works once and expires after 15 minutes), then checks when it lands.

  • fetch_document — retrieves a stored file over the signed-in connection.

Shared spaces

  • list_shared_spaces, list_space_memories — see your spaces, who's in them, and everything shared into a space.

  • open_shared_space — creates a space and can ask to invite people by email as viewers, including people who don't have an account yet.

  • share_memories — asks to share memories into a space, or takes a share back.

  • close_shared_space — archives a space you own and ends everyone's access. You can reopen it in the app.

Sharing and inviting need your approval. Your assistant gives you a link to app.justonce.ai/confirm/…; nothing is shared and nobody is invited until you approve it there, and the link expires after 24 hours. Taking a share back and closing a space happen straight away. Creating a shared space needs a paid plan.

ChatGPT and Codex

  • search and fetch — the citable-results tools ChatGPT connectors use. Other assistants use search_memories and get_memory instead.

  • ChatGPT and Codex don't get get_core_profile, get_context, get_as_of, get_relationships or find_related_entities, and some memories are kept out of them. See Connect ChatGPT.

Writing tools need the "save, update and delete your memories" permission you allowed when you connected. When your monthly allowance is used up, write tools tell your assistant why instead of failing silently. Searching and reading never use your allowance.

Did this answer your question?