How Gryz search finds your documents, notes, memories and tasks, who sees what, where your text goes, and how to turn it off.

Search is rolling out. It may not be on for your account or your company yet.

Answer first

Search lets you, and the AI tools you connect, find the right passage in your own Gryz documents, notes, saved memories and tasks, or in your company's documents, memories and tasks. To make that work, Gryz splits your text into short passages and sends those passages to Cloudflare, our AI provider, to prepare them for search. You can turn search off at any time, and you can leave out any document, note or project. Nobody gets results from anything they could not already open.

What gets searched

Your documents and notes. Gryz searches the text of documents you wrote and, in your personal space, your notes.
Your company's documents. In a company workspace, search covers the company's documents. An owner or admin has to turn it on first.
Saved memories and tasks. Search also looks through the memories and tasks you could already open, in the same space, and labels each result so you can tell a document from a memory or a task. An AI tool only finds memories and tasks it could already look up on its own: a team member never finds your personal memories. Tasks are matched by the words in their title and description. You can switch memories and tasks off in Settings.
Only text. Markdown and HTML documents are searchable. Images and other file types are not.
Not everything. These are never made searchable: documents with an expiry date, documents protected by a PIN, documents held for review, and anything you leave out yourself.
Long documents. Very long documents are searchable up to a limit, so only the first part may be included.

Who sees what

Your private space is yours. A search in your private space only looks at your own documents, notes, memories and tasks. It never returns company content.
A company search stays in that company. It never returns your personal documents, memories or tasks, or anything from another company.
People only get what they can already open. In a company, a result comes back only if that person could open the document, memory or task anyway. What sits in a restricted project stays inside that project's people.
Agents follow the same rules. A connected AI tool searches as you, with the same limits. A company agent only reaches what it has been given access to.
Access changes apply at once. If someone loses access to a project, their next search no longer returns it.

Where your data goes

Your text is sent to Cloudflare. To prepare a passage for search, Gryz sends its text to Cloudflare, which turns it into a list of numbers that captures its meaning. Cloudflare is listed as a sub-processor in our privacy policy.
The same goes for your questions. The words of a search are sent to Cloudflare so they can be matched against your passages.
Memory already works this way. The text of the memories you save is already sent to Cloudflare for the same reason. Searching your memories sends only your question, the same way looking up a memory always has. Tasks are matched by their words inside Gryz and are not sent anywhere.
Cloudflare says it does not train on your content. We have not found a formal promise that Cloudflare deletes the text immediately after processing it.
The results are stored in Gryz. The passages and their numbers are stored in our own database, not with Cloudflare. They are derived from your documents and can be rebuilt, and they are deleted when the document, project or account they come from is deleted.
We do not record your questions. Gryz keeps counts so we can watch cost and reliability. It does not keep the words you searched for.

Your controls

Turn search off. In Knowledge, open Settings and switch off "Make this searchable". Gryz deletes your searchable passages and stops preparing new ones. Your documents and notes are not touched. Turn it back on and Gryz prepares them again.
Company default is off. A company owner or admin has to switch search on for the company, after a note that text is sent to Cloudflare. Turning it off deletes the company's searchable passages. The change appears in the company Activity page.
Choose what search includes. Documents are always included. You can switch notes, saved memories and tasks on or off for the space. This only narrows search; it never shows anyone more.
Leave things out. On the same page you can leave out single documents, notes or projects. Their passages are deleted right away, and they stay out until you put them back.
Choose how strict search is. Standard returns anything related, so you rarely miss an answer. Strict returns fewer, closer matches.
Re-check results (experiment). An optional second step re-orders the best matches. It is off by default because it did not make results better in our tests.
Try a search. The Knowledge page runs the real search as you, and shows why each result came back, so you can see what your agents would get.

What "no source found" means

When search finds nothing close enough, it says "no source found". That only means nothing in the space is even about the topic. It does not prove your documents do not answer the question.

Search can still return a passage that is about the right topic but does not contain the answer. In our tests, on questions whose answer was missing from the documents, standard search still returned a related passage most of the time. Read the passage before you rely on it, and treat a weak match as a lead, not an answer.

What stays private

Your personal documents and notes are never readable by your company, its owners or its admins, and they are not part of any company search.
Search does not make anything public. Results appear only to the person, or the agent acting for them, who could already open the document.
Documents protected by a PIN and documents with an expiry date are never made searchable.

How search works and where your data goes