The paper you read in May. The connection you make in October.

Picture a wall of archive boxes, one per paper, each with a label. Every time you read a new paper you put a fresh box on the shelf with the title on the front and your own notes about the paper folded inside. An archivist walks behind you, reading each new box, and quietly attaches a coloured ribbon between any box and the older ones it talks to (same method, same author, same dataset). Three months later you reach for a new box and find five other boxes already tied to it.

Concretely: drop a PDF into a Korely note. The text gets extracted into the body, the file stays in the attachments/ folder of your vault. Add your own summary above (the authors, the claim, why it matters to your work). The entity extractor picks up named methods such as GraphRAG or contrastive learning, plus authors and datasets, automatically. When the next paper arrives the related notes panel already shows what it connects to. More on the entity extractor →

Picture the senior librarian at a small academic library. You walk in and ask "which papers in my own reading propose alternatives to softmax attention?" They walk to the shelf, pull the four boxes that match, hand them to you, and tell you which boxes the quotes came from. Korely's MCP server does the same thing for Claude, Cursor, and any other MCP-capable assistant.

Writing a literature review chapter? Ask the assistant to draft the section, pulling from your own notes. Drafting a grant proposal? Same move. Only the relevant excerpts go into the model's context window. The notes themselves never leave your machine on Free, and only leave through the cloud MCP endpoint you have explicitly turned on with Pro.

Picture the corner of your office where your private reading shelf already lives, with the papers you actually annotate, the drafts you are still polishing, the preprint you are not ready to share yet. Free Korely is the index card system that sits next to that shelf, on the same desk. Three useful properties come from running on your own machine.

• Search and graph work offline. On a plane, on a train, in a library with no wifi, your whole reading is one keyword away.
• It plays well with your reference manager. Tools like Zotero and Mendeley keep doing what they are good at (citations, metadata, BibTeX export for your papers). Korely sits next to them as the place where your thinking about the papers lives.
• It fits research environments where data residency matters. A university IT policy that is cautious about cloud note apps has nothing to flag with Free Korely: there is no cloud component to turn off.