Knowledge vault

The knowledge space is built on a fork of Rowboat, an open-source knowledge-graph application. It holds knowledge notes, drafts and briefings, and the ideas space syncs into it. A headless instance runs on our own server; a desktop part runs on a workstation.

Plain files on purpose

Knowledge notes are kept as files in a vault directory, readable in any editor. This is a deliberate constraint. A knowledge base you can only read through the product is a knowledge base you can lose. Files survive the product being abandoned, including by us.

On top of that sit the application’s own document store and vector index for search. If the derived layers are lost, they rebuild. If the files are lost, nothing rebuilds — so the files are the source of truth.

What it is used for

• ·Search across captured knowledge
• ·Reading and writing knowledge documents through a small MCP server
• ·Drafting and briefings with the vault as source material rather than the model’s memory