Kanwas: A shared board where your AI agent's tool calls live in public

What it does

Kanwas is a real-time collaborative canvas for teams working with AI. Humans drop documents, notes, and decisions onto a shared board, while an AI agent operates on the same surface—its tool calls and reasoning stream into a shared timeline visible to everyone. A companion CLI syncs the workspace to a local git-backed markdown filesystem, so you can edit in your own editor or feed specs straight to Claude Code or Codex.

The interesting bit

Most AI tools hide the agent’s reasoning in a chat sidebar; Kanwas makes the agent’s mechanics part of the shared workspace. By backing everything with a git-tracked markdown filesystem, it also treats your work as plain text that compounds rather than proprietary data that rots.

Key highlights

• Real-time multiplayer canvas built on Yjs and BlockNote.
• AI agent shares the same board, with tool calls appearing in a transparent timeline.
• Git-backed markdown export with full version history; no vendor lock-in.
• CLI (kanwas) for local sync, bulk markdown import, and non-interactive CI scripting.
• Self-hosted stack requiring Docker and your own LLM API keys.

Verdict

Worth a look if your team is tired of context-switching between chatbots, docs, and task trackers and wants a single shared surface for AI-assisted work. If you need a fully managed, zero-config SaaS, the repo is aimed at self-hosters.