Asking Your Agent To Search Everything You Can See
Ask one question and have your agent search by meaning across your files, notes, tasks, projects, memories and past conversations - including everything shared with you.
Overview
Your agent can search across everything you have access to in one question — your files, your notes and pages, your tasks and goals, your projects and plans, your saved memories, and your past conversations — and it searches by meaning rather than by matching words.
That distinction is the whole point. If you ask "what did we agree about the renewal?", a word-matching search only finds material that literally contains "renewal". A meaning-based search finds the page where someone wrote "we'll extend for another twelve months at the same rate" — which is the answer, and which shares not one word with your question.
It covers what other people have shared with you as well as what you own. That half is usually the hardest to go looking for by hand, because you did not create it, you may not remember who sent it, and you have no idea which project it ended up in.
How to use it
There is nothing to turn on and no new button. You ask your agent a question in the ordinary way, and it searches when searching is the right way to answer you. Questions that work well:
- "Search everything I have access to — what did we agree about the Q3 renewal?"
- "Find that conversation I had a while back about the pricing model, and quote it."
- "What do my notes and my past conversations together say about the Northwind project?"
You do not have to know where the answer lives. That is the part this replaces — the "which project was that in again?" step you would otherwise do first.
What comes back
Answers cite the object they came from. When your agent tells you something it found, it names the note, file, task, project or conversation behind the claim, so you can open the original and read it in context rather than taking the summary on trust.
A useful habit: if an answer matters, ask "which object did that come from?" and open it. The citation is there precisely so a claim can be checked.
What it searches
Type | What that covers |
Files | Documents you have uploaded, and documents shared with you |
Pages / notes | Your written notes and pages |
Tasks | Tasks and subtasks, including their detail |
Goals | Goals, wherever they sit in a project |
Projects | Project descriptions and detail |
Plans | Plans you or your agents have written |
Memories | The things Sapience has remembered about how you work |
Conversations | Whole past conversations — searched as complete conversations, not as loose individual messages |
Conversations being searched whole matters in practice: a decision made across five messages of back-and-forth is findable as one thing, which is how you actually remember it.
It searches as you, not as the agent
This is the part worth understanding, because it is what makes it safe to use on real work.
The search is scoped to you, the person asking — not to the agent you happen to be talking to. Two colleagues using the very same agent each search their own material, and neither can reach the other's. The agent is the thing you are talking to; it is not the thing that decides what you are allowed to find.
Practically, that means you can use any agent that offers this without wondering whether you are seeing someone else's documents, or whether they might see yours through the same agent.
What happens behind the scenes
- Your material is prepared for searching in the background — you never trigger this, and there is no "index now" button to press.
- When you ask a question, your agent decides whether searching would help answer it.
- If it does, your question goes to the search as a meaning query rather than a keyword one.
- The results come back with the object each one came from attached.
- Your agent reads them, answers you in plain language, and cites the objects it used.
New material becomes searchable on its own, shortly after it is created or shared with you. There is nothing to press and nothing to wait for deliberately.
Tips
- Ask the way you would ask a colleague. "What did we decide about X?" beats a keyword list — meaning-based search is at its best with a real question.
- Say the type if you want to narrow it. "Search my notes and pages for…" is a useful steer when you already know the shape of what you are looking for.
- Follow the citation. The named object is the original; the answer is a summary of it.
- Recent material may take a short while. Something created moments ago may not be searchable in the same minute.
Summary
You can now ask one question and have your agent search across everything you have access to — your own material and everything shared with you — by meaning rather than by keyword. Answers cite the object they came from, so you can always open the original. The search is scoped to you as the person asking, never to the agent you are talking to, so the same agent gives two colleagues two different, correctly-separated sets of results.