The point of a vault is ownership. It’s just files and folders on your machine or in your synced storage, which means the business knowledge has a home that survives when someone leaves the company or a vendor changes software. You own the files.
Think about an office filing cabinet you actually own. Client folders, project folders, and reference notes all go in it. Unlike a paper cabinet, though, one note can connect to many others without copying the same text everywhere. That’s what makes it a working system instead of a static archive.
How it shows up
In AI work, a vault becomes the place where context compounds. A raw source comes in, a meeting note gets processed, a project status gets updated, a decision note preserves why something changed. Over time it becomes a knowledge base both people and agents can read. Most notes are written in markdown, plain text that AI tools read easily, so your company memory isn’t locked inside a weird export format. Notes point to each other through a wikilink: a client links to a project, a project links to a decision, a decision links back to the meeting where it was made.
Why you care
When you use Claude Code, Codex, or another agent with a vault, the agent can inspect the files, follow the links, and pick up context from the system you already maintain. That makes the model less dependent on one giant prompt, and it keeps your context yours instead of trapped in a single chat window.