Warren is a fast, native workflow for AI coding. Keep Project, Workspace, Session, Runtime, and Agent in one model. - Host-owned Sessions that survive client quits and reconnects - A single Project → Workspace → Session resource model - Raw Terminal and structured Agent views - Native macOS terminal UI with libghostty; libghostty-vt runtime; xterm.js on Web - macOS, Web/PWA, mobile Web, and CLI clients - Open source under Apache-2.0 Learn more:
I built Warren to make AI coding feel like one workflow instead of scattered windows.
The core is the relationship between Project, Workspace, Session, Runtime, and Agent.