Spatial Agent Command Center
Spawn, monitor, and voice-command teams of AI coding agents in an immersive spatial computing environment
Why existing tools fall short
Andrej Karpathy showed us the future: teams of AI agents coding in parallel. But managing them means squinting at flat tmux grids, switching between terminal tabs, and losing track of costs and progress. As agent swarms grow, 2D interfaces become a bottleneck.
SACC transforms this into an immersive 3D command center where you can see all your agents at once, speak commands naturally, and manage tasks spatially — purpose-built for Apple Vision Pro and the WebSpatial platform.
Everything you need to command your agent swarm
Live tmux output with full ANSI color rendering, team grouping, and click-to-maximize panels
Kanban board powered by Claude Code's native task JSON files with drag-and-drop status updates
Speak commands to agents and hear summaries with distinct ElevenLabs voices per agent
Per-agent cost tracking, token usage, model identification, and aggregate team metrics
A-/A+ controls optimized for Apple Vision Pro readability at spatial computing distances
Fuzzy-matched project groups with color coding for organizing agent swarms
Two-process system connected via WebSocket
Spatial Computing Interface
React + Zustand + WebSpatial · port 3333
FastAPI + watchfiles · port 8765
tmux sessions + Claude Code task files
The bridge server watches tmux sessions every 2s, captures terminal output with ANSI escapes, parses agent stats from Claude Code status bars, and syncs task files — all streamed over a single WebSocket connection.
Modern tools for a modern interface
Designed for spatial, works everywhere
Primary target via WebSpatial SDK — spatial windows, glass panels, and elevated UI elements
Fully responsive fallback — all features work in Chrome, Safari, Firefox
SenseAI WebSpatial Track