49Agents launched on GitHub this week calling itself the first '2D agentic IDE.' The open-source project replaces scattered terminal tabs with an infinite canvas where you arrange terminals, editors, and AI agent sessions spatially. Zoom out, see everything at once. Zoom in, focus on one task. Running AI agents across multiple machines is a pain. SSH connections everywhere. Terminals you lose track of. 49Agents connects all your machines through a WebSocket relay. No SSH required. You can pull up your workspace from a phone, with live CPU, RAM, and Claude API usage displayed across each connected machine. Under the hood, each machine runs a 49-agent process connecting to a relay server via WebSocket. Terminal sessions use real tmux through ttyd. Full ANSI color. Full scrollback. The relay self-hosts or runs through 49agents.com. Privacy is straightforward: terminal I/O is relayed but never stored server-side. There's also Beads integration, a distributed graph issue tracker for AI agents built on Dolt (a version-controlled SQL database). Agents get structured memory and dependency-aware task management. The code ships under BSL 1.1, free for individuals and small teams, converting to MIT in 2030. The zero-SSH multi-machine setup alone makes this worth trying for anyone juggling agents across several servers.