Orchestrate AI agents locally from your ideas, notes, and research.
Each sandboxed in its own worktree.
Powered by your existing Claude Code subscription.
How It Works
A compounding loop between what you know and what AI creates. Agents research, build, and feed discoveries back into your knowledge.
Your knowledge feeds the agents.
Ideas, notes, and research live as local files you own. This knowledge becomes the foundation that shapes how agents think, research, and create.
Run safely in worktrees.
Each agent runs in its own git worktree. They can research, write code, generate assets, and update files without touching your main branch until you're ready.
Create in parallel.
Launch multiple agents from a single task screen. Each agent gets its own worktree sandbox and timeline, letting you run experiments in parallel.
Curate in one editor.
Flip through every agent’s commits, logs, and assets inside the unified knowledge editor, diff changes inline, and merge only the pieces you approve so the main workspace stays pristine.
Get Started
Free during beta. Local-first, works with your Claude Code subscription.
