About this project

Xum (formerly Mux) is a desktop application described as a coding agent multiplexer, designed for isolated, parallel agentic development. It lets you manage a suite of coding agents at once, each in its own workspace, with a central view of git divergence. Key capabilities evidenced in the README: - Isolated workspaces with three runtime options: Local (run directly in your project directory), Worktree (git worktrees on your local machine), and SSH (remote execution on a server over SSH). - Multi-model support, listing sonnet-4-*, grok-*, gpt-5-*, and opus-4-* models, plus Ollama for local LLMs and OpenRouter for a broader range of models. - A VS Code extension to jump into Xum workspaces from the editor. - UI and keybinds for managing multiple agents efficiently. - Rich markdown output including mermaid diagrams and LaTeX. - A custom agent loop with familiar features such as Plan/Exec mode, vim inputs, /compact, plus opportunistic compaction and mode prompts. - Integrated code review, agent status reporting in a sidebar, git divergence UI, cost and token consumption tracking, and a context management dialog. - A server mode with a responsive UI for mobile users. Installation is via pre-built binaries from the releases page for macOS and Linux. The project is licensed under AGPL-3.0. Documentation is hosted at mux.coder.com, and development setup is described in AGENTS.md. The project was renamed from Mux to Xum following a trademark concern raised by Mux.com.