Hey PH I'm Toshal, and I just open-sourced loomcode an MIT-licensed AI coding agent that lives in your terminal instead of a browser tab.
The idea started from being tired of switching between chat windows and my editor. So loomcode plugs into whatever model you already pay for Anthropic, OpenAI, Gemini, OpenRouter, NVIDIA NIM (free tier), or local Ollama and you can switch models mid-session with /models. No lock-in, your own keys.
The part I'm most unsure about (and would love feedback on) is Loom Computer, a beta feature where the agent can actually see your screen and control it clicking, typing, switching windows through one floating pill. I added a panic hotkey, password-manager blocking, and action caps, but I know giving an agent "hands" on a real desktop is a trust question, not just a technical one.
Loom Computer is in beta and it grew up fast. The agent now works like a human at the keyboard: it looks, thinks, acts,
then VERIFIES the screen changed and switches strategy instead of
repeating failed clicks. It narrates every step, shows a blue ripple
wherever it clicks, and finishes with a verified summary. New since beta start: One floating pill UI input, model picker (all providers, searchable), pause/stop/quit. Grows upward, never pops up while you work. Vision AND non-vision models non-vision gets a text map of the screen Subagents fan out non-GUI subtasks of big jobs Background mode shell/files/web only while YOU use the PC 8x faster steps (compressed screenshots) Safety: Ctrl+Alt+Q panic, password-manager blocking, action caps, full audit trail, resume sessions Still coming: TUI control, web API routes, UIA pixel-free clicking,
Wayland, per-app allowlists. This is for Beta only Try the beta: npm install -g loomco loom computer
For trying normal stable production build use: npm install -g loom-agent
