Knowledge and Search

Your Knowledge is what Kojo can draw on: the knowledge built from all your meetings, plus documents. Under the hood, every meeting is processed into a knowledge graph -- see GraphRAG Memory Engine for the full technical architecture.

The Knowledge page

• Desktop: Sidebar > Knowledge (/dashboard/memory).
• Mobile: Bottom nav > Knowledge (/m/memory).

The page opens on Documents. It has three tabs on desktop:

• Documents, upload PDFs, Word and Excel files, images, text and code files, ZIPs, and audio recordings, with your storage usage alongside.
• Knowledge, coming soon.
• Sources and controls, which meetings feed the workspace's memory (and moving one between your personal space and the workspace), and a summary of what Kojo can use.

Mobile has the same three tabs, in the order Docs, Knowledge, Sources, and also opens on Docs. Docs lists your documents with their status and shows your storage usage. Uploading documents (including audio) is desktop only for now, and uploaded files are searchable from mobile once they're indexed. You can browse folders and change a document's audience from mobile.

Meetings, action items, decisions and past Kojo chats each have their own page in the sidebar, so this page no longer duplicates them.

What Kojo can draw on

Kojo only ever answers from meetings and documents you have access to. A meeting scoped to a group you are not in, or restricted to specific people, never appears in your answers, your lists, or your search results. See Workspaces and Teams for how that is set.

Knowledge tab

Shared, promotable organizational knowledge is coming soon. The Knowledge tab shows a notice until it ships.

Documents

Upload PDFs (including scanned ones), DOCX, Excel workbooks (XLSX), images (PNG, JPG, WEBP), TXT, MD, CSV, JSON, ZIP files (extracted automatically), common code files, and audio recordings (mp3, wav, m4a, webm, ogg, aac, flac). Click Upload or drag files onto the page, then choose your personal workspace or the team workspace as the destination. The workspace you are working in is selected for you, and the file is saved there immediately, visible to everyone in that workspace who can see it.

Files and images you attach in a Kojo chat can be kept the same way: click Save to memory on the attachment in the chat.

Audio recordings are transcribed automatically and become searchable the same way a document does. Unlike other document types, a recording has a duration cap (your plan sets the maximum length for a single upload) and a small transcription charge of 1 credit per minute, shown as an estimate before you confirm.

How Kojo reads your files

When you upload a file, MemorDesk stores its readable text and, for files that have a structure, an outline of what the file is: a JSON file becomes a description of its shape and names (an animation file is recognised as one, with its size, length, layers and controls; a package.json, an API description or map data are recognised too), a CSV becomes its columns and types, code becomes its functions and classes, and Markdown and web pages become their headings. HTML is read as the text you would see on the page.

• Ask Kojo about a file. Desktop: Knowledge > Documents, open a file's card and choose Ask Kojo about this. The file is attached to the chat as a chip; press send. Mobile: Knowledge > Docs, Ask Kojo on the file. Kojo reads that file first, in full when it fits or its outline, opening and the parts that match your question when it is long, and only then connects it to your meetings and other documents.
• Scans and pictures. A scanned PDF (one with no selectable text) and an image (a photo of a whiteboard, a receipt, a screenshot) are read by AI when you upload them: every page or picture is transcribed, with tables as rows, and a picture also gets a description. Kojo tells you that text was read by AI, because handwriting and poor scans can contain small mistakes, so check important figures against the original. Pictures inside a ZIP are not read; upload them on their own.
• Excel workbooks. Every sheet is read cell by cell, with the value each formula last calculated, and dates shown as dates. Kojo sees each sheet's columns and rows, so questions like "what was the total for Lagos in March" are answered from the actual cells. Charts, pictures and comments in a workbook are not read, and the older .xls format is not supported (save it as .xlsx).
• Name a file. Naming a file in your question ("what is in Budget 2026.xlsx", or its name in quotes) works too. You can also type @ to pick a file or meeting.
• Follow-up questions keep reading the same file for the next few turns, without attaching it again.
• See what Kojo understood. On a file's card, How Kojo reads this shows the type and outline, so you can check Kojo has understood the file correctly. A file uploaded before this feature is read in full the first time you ask about it.
• Long files. Kojo reads them in parts, and asks for the next part or the part matching a topic when it needs it. It tells you when it has only seen part of a file.
• Access. Kojo only reads files you can already see. A file in a folder or group you are not part of, or marked private, is never read for you.

Folders and who can use a document

In a team workspace, documents can be organised into folders, and every folder and document has an audience: Everyone in the workspace, Specific groups, or Only me.

• The folder is a ceiling. A document inside a folder can only ever be as open as the folder. A folder for one group keeps everything in it to that group, and a folder inside it can only narrow that further, never widen it. Folders nest up to five levels deep.
• Your default. New uploads start with the workspace default, which owners and admins set in Settings > Workspace under Who can use new documents (everyone, the uploader's groups, or only the uploader). You can pick another audience for a single upload in the upload dialog. Chat exports always start as Only me.
• Confidential. A document marked Confidential stays out of Slack answers and shared answers in every channel, whoever is in it, and is never sent by an integration.
• Sharing with one person. From a document's Audience dialog you can give one workspace member access for a chosen time (up to a year), even when the folder would not otherwise include them. The same works for a whole folder. This needs a plan with Groups.
• Moving a document. Moving one into a folder that admits more people asks you to confirm first.
• When someone leaves. Their team documents are handed over to the workspace owner, and the owner is notified.
• Who can change what. The uploader, workspace owners and admins can change a document's audience and folder. Folders are managed by whoever created them, plus owners and admins.

Desktop: Knowledge > Documents. A folder bar sits above the list with a New folder button and a Shared with you list. Each document has an Audience button, and a small label appears only when the audience is not the default.

Mobile: Knowledge > Docs. The same folder bar appears above the list, and each document has an Audience button that opens the same dialog. Uploading is still desktop only.

Global search

Search across meetings, action items, decisions, and notes.

• Desktop: the search bar in the centre of the top bar.
• Mobile: the search icon in the top-right pill of the top bar.

What is indexed

Control what is searchable in Settings > Knowledge (desktop) or Settings (gear) > Knowledge Base (mobile): meeting transcripts, action items, decisions, and cross-meeting insights.

Memory window by plan

How far back Kojo can draw from when answering questions:

Meetings outside your window remain stored but are excluded from Kojo's context and search results until you upgrade.

Opening a meeting outside your window. On desktop the meeting page opens, with the summary blurred behind a card that says "This meeting is outside your memory window", how many days your plan covers, and an Unlock with Team (or Unlock with Pro) button. The other tabs are greyed out. On mobile the same card fills the Participants tab, and Summary, A. Items and Transcript are disabled. Upgrading, or moving to a plan with a longer window, makes the meeting readable again.

How search works

MemorDesk uses vector similarity search: each paragraph of each meeting is converted to a 2048 dimensional embedding. Search returns results ranked by semantic meaning, not keyword overlap. You can search for concepts ("budget concerns from last sprint") and get matches even when the exact words do not appear in the transcript.

Embeddings are contextualized, meaning each paragraph is encoded with awareness of the paragraphs around it rather than in isolation. A line like "we agreed to push it back" keeps the sense of what "it" referred to earlier in the meeting, so it can still be found later.

If the vector index is ever unavailable, search falls back to keyword matching over the same text rather than returning nothing, and Kojo says when grounding is limited.

Each user's search index is partition-isolated -- queries can never cross into another user's data even under simultaneous load.