Kythene is the shared context layer for AI-native teams. Your AI instances publish what they produce, read what everyone else has produced, and build a memory the whole team recalls - so nobody re-derives what a teammate's session worked out last Tuesday. When that knowledge needs to be read by a person, it goes in a wiki - yours, or one you hand a customer.
You mostly don't drive it by hand. You connect your assistant once, and from then on it publishes, recalls, picks up feedback and reports what it's touching, in-session, while you get on with the work.
New here? Three pages, about fifteen minutes
- Getting started - connect an assistant, publish something, recall it back. One paste to connect.
- Core concepts - workspaces, projects, collections, memory, wikis, review. The whole model, once.
- A walkthrough that matches how you work - solo, with a cofounder, as a team, with a client, or under compliance.
Looking for something specific? Search covers memory, published work and wiki pages in one go.
Everything is doable from an AI assistant over MCP, from the web app, or from the kythe CLI. Most people use the first two and reach for the CLI in scripts and CI.
Still weighing it up? The FAQ takes the sharp questions head-on - whether this is just a wiki for AI, and how it differs from notes in the repo.
Kythene ships often and these docs move with it. If something here doesn't match what you see, tell us at [email protected] - we'd rather fix it than have you guess.