Sources

The file layer behind CoachKeeper — upload PDFs, Word docs, images, and text files, attach them to notes, and let the AI assistant and your connected agents read them.

What are sources?

Sources are the files behind your workspace: PDFs, Word documents, Markdown, text files, and images. They’re stored as-is and are viewable, downloadable and attachable the moment the upload finishes.

One thing keeps working in the background afterwards: CoachKeeper reads the text out of your file so it can turn up in search and in what the assistant cites. The card says Indexing… while that runs and clears itself when it’s done — you don’t have to wait for it, and nothing else about the file depends on it.

Sources isn’t a tab you live in — it’s a supporting layer. You’ll mostly meet your files where you already work:

• Attached to notes — files travel with a note as chips, one click from the built-in viewer
• In the Library — tagged files appear as cards on their tag pages, next to the notes and todos that share the tag
• Read by your agents — the assistant reads a file when you mention it by name, and any agent connected via MCP can list your files and read their contents directly

The Sources view itself is the file manager behind all of this — upload, folders, renaming, bulk actions. Open it from the command palette (Go to Sources) or the /sources route.

Attaching files to notes

The most common way to use a source day to day:

1. In the editor, click the paperclip button below the note
2. A popover browses your file library — navigate folders or type to filter — and a click attaches the file
3. The file appears as a chip on the note; click the chip to open it in the built-in viewer, side by side with your note on desktop

Removing a chip only unlinks the file from the note — the source stays in your library. The paperclip picks from files already in your library, so upload new material first.

This keeps a note and its supporting material together: a meeting note with the agenda PDF attached, a literature note pinned to the paper it discusses.

Uploading a source

1. Open Sources (command palette → Go to Sources)
2. Click Upload and select a file (PDF, DOCX, TXT, or MD)
3. That’s it — the file is stored and immediately ready

You can also drag and drop files directly onto the Sources view.

Size limits: 50 MB for documents, 10 MB for images, 5 MB for plain text files.

Looking for saved links?

Web links aren’t files, so they don’t live here. Paste a URL into Quick Capture and it becomes a Link — a first-class capture with its own title, comment, and tag — collected in the Library, where every link also joins an automatic shelf for its website.

Supported file types

Note: text extraction works on PDFs that contain a real text layer. Scanned (image-only) PDFs can’t be read yet — the file uploads and displays fine, but it’s marked Not searchable and won’t appear in search results or agent reads. That badge is only ever about searching: the file itself is stored and opens normally.

Reading sources in the app

You don’t have to leave CoachKeeper to read what you uploaded. Open a file — from a note chip, a Library card, or the Sources view — and it appears in the built-in viewer:

• PDFs render inline with full page navigation
• Markdown and text files display as formatted text
• Images open in a lightbox
• Word documents and other formats show a file info panel with a text preview when available — and a download button when in-app preview isn’t possible

Files your agents can read

Sources aren’t just for you. Connect an agent over MCP — Claude, Cursor, any MCP client — and it can list your files, look them up by name, and read their full text. Upload a spec and your coding agent can read it while it works; upload a meeting transcript and your assistant can turn it into tasks.

The built-in chat works the same way: mention a file by name and the assistant reads its content to answer your question — no indexing step in between.

@Mentions in chat

Type @ in the chat input and select a source to pin it to the conversation. The AI gets the full content of that document as context — useful when you want to ask detailed questions about a specific file without the AI having to look it up.

Organizing your library

• Folders — create folders and nest them to organize your documents by project, topic, or category
• Drag and drop — move files between folders by dragging
• Grid and list views — switch between visual grid and compact list layout
• Rename — click a file or folder name to rename it inline
• Multi-select — hold Shift and click to select multiple items for bulk actions

Use cases

• Book reviewers — upload the book PDF, then mention it in chat and ask the AI to help you draft a review based on your highlights
• Researchers — upload papers and have your connected agent read and compare them
• Students — upload lecture slides, then name the file and ask the AI to create study todos based on the material
• Knowledge workers — attach meeting transcripts, contracts, or strategy docs to the notes that discuss them

File management

Each source shows its filename, file size, and upload date. You can delete sources you no longer need; notes that referenced a deleted file simply show a removable “missing file” chip.

For large files, a confirmation dialog appears before upload to let you know the upload may take a moment.

Storage limit. Uploads and images you paste into notes share 1 GB per account on the hosted version. Notes, todos and events don’t count toward it. You can see what you’re using at any time in the avatar menu under Usage; if an upload would go over, it’s refused with a message telling you how much room is left.

Tips

• Build your library as you go. Anything you upload is instantly ready — for you in the viewer, and for the assistant or your agents when you point them at it.
• Attach, don’t hunt. If a note discusses a document, clip the file to the note — future you gets the material one click away instead of digging through folders.
• Use @mentions for focused questions. If you want to ask about a specific document, pin it with @ — the assistant gets the full file as context.
• Organize with folders. A flat list of 50 files is hard to navigate. Group by project or topic for quick access.