← Back to blog

How to Ask AI Questions About Your Own Notes Safely

A practical way to ask useful questions about saved notes while checking sources and understanding what Zenodot's read-only MCP beta can access.

An assistant can help you find patterns in your notes, but the useful question is rarely “What do my notes say?” A better question names the task and asks for the saved evidence behind the answer. Before connecting an outside AI client, you should also understand what it can read and how to remove access.

Ask about a decision, not your entire Library

Imagine you are revisiting a product decision from last month. You remember two options but not why you chose one. Ask:

“What trade-offs did I record between the two onboarding approaches? Point me to the Pages behind each reason.”

That request gives the assistant a purpose and asks it to show its work. Open the relevant Pages after receiving an answer. Check whether the cited material actually supports the conclusion, and look for a missing or newer note before acting on it.

If you have no saved evidence on the question, a confident-sounding answer cannot create it. Make the gap explicit and gather the missing information yourself.

Know which surface you are using

In Zenodot’s app, you can chat with the Librarian about your saved Library. The separate Zenodot MCP beta lets a compatible outside client search and read the current Library for a linked Pro account. Its current tools can search Pages, read a Page, browse Page or Folder metadata, and follow explicit Page links. The beta is read-only: it cannot edit the Library. It also does not expose raw Chute Notes, Page history snapshots, or your conversations with the Librarian. The capabilities guide lists the current surface.

Read-only describes what the MCP server allows a connected client to do to Zenodot. It does not mean the outside client cannot receive or process the Page content it requests. Review that client’s own privacy settings and policies before connecting sensitive material.

Treat the connection key like a password

Zenodot MCP keys grant library:read access to the linked account’s current Library. Create a key only for a client you trust, choose an expiry that fits your use, and never paste the key into a prompt or public document. Revoke it when you stop using the connection. Zenodot’s privacy and access guide explains key scope, expiry, and revocation; the connection guide covers setup.

For particularly sensitive notes, decide whether they belong in a connected Library at all. A useful answer is not worth sharing information with a service you have not evaluated.

Check the source before you use the answer

An AI response is a starting point for retrieval. Open the cited Page, check its date and context, and compare it with any newer decision. Then record the action you will take: continue a draft, revisit a choice, or ask someone for missing evidence.

This is the same discipline that makes research notes useful: preserve the source, distinguish it from your interpretation, and return to it when the stakes rise. Download Zenodot for iPhone if you want to try the in-app workflow; MCP access currently requires a linked Pro account and a compatible client.

Topics

  • AI note taking
  • note privacy
  • MCP