Klawforge Philosophy

  1. Klawforge takes the natural-language exchange already happening between you and your AI tools — and treats it as the source of knowledge, skills, conventions, and workflows.
  2. No new syntax to learn. No special tags like @codebase, #file, or @workspace. You keep chatting with Claude / Codex / Gemini the way you already do.
  3. In the background, Klawforge captures what is worth keeping, sublimates it into a graph, and quietly surfaces context when your next session needs it.
  4. The product gets smarter as you use it. Tools feel like they remember. Memory travels with you across IDE, session, model.
  5. And eventually you can pick one branch of your graph and share it — not your whole brain, just the subtree that matters — with another OneBoss.

Five layers, all from one paragraph

  • Input — natural language only; never force the user to invoke tools with a special syntax.
  • Processing — sublimation across knowledge, skills, conventions, and workflows.
  • Surface — IDE-side quiet enhancement; the model behaves smarter without you doing anything.
  • Boundary — environment and model abstraction; you do not pay twice for the same compute.
  • Sharing — surgical subtree share, not whole-brain dump; the network is horizontal-mesh, not centralized.