MudrikNow — Open-Source
Stop explaining your screen to AI.
Your AI sees what you see.
MudrikNow is an open-source desktop AI agent that sees what you see — and answers, acts, or guides you step-by-step through any task.
Why MudrikNow
Three promises.
AccessInstant Access
Show MudrikNow instantly with a shortcut, whenever you need it.
ContextLess Context
It understands what's on your screen, so you explain less.
ReachAny Task. Any Window.
Get help and guidance through any task and any window.
Demo
See it in action.
Bring your own model
How it works
Install in four steps.
From zero to an AI that sees your screen in about two minutes.
Install the AI runtime
One global npm package. Open PowerShell and run:
Run the MudrikNow installer
Download the installer and double-click.
MudrikNow 3.3.0 installer (.exe) →Add your AI model
Open ⚙ Settings → API Keys → paste a provider key (Anthropic, OpenAI, Google, Kimi, DeepSeek…), then pick a model. A multimodal (vision) model is recommended — MudrikNow auto-screenshots every activation; Google Gemini Flash-Lite is a great free pick. Or authenticate via CLI:
First launch triggers Windows SmartScreen — More info → Run anyway.
What it does
Six ways MudrikNow goes beyond chat.
VisionSees any app
Reads the full UI tree — buttons, fields, text, menus — through native accessibility (UI Automation). A full-screen screenshot with a coordinate grid is captured on every activation.
SandboxedClosed by default
Read-only shell for diagnostics (writes, deletes, piping blocked — violations kill the session). No filesystem writes. The model runs in a sandbox with an allow-listed capability surface.
GuideAuto-Guide
MudrikNow walks you step-by-step through multi-step tasks. An owl cursor points to each target with instructions in a speech bubble.
ChatQuick chat
Press Alt+X — the panel opens instantly without scanning your screen. Always one keystroke away.
ModelsBring your own model
Anthropic, OpenAI, Google, Groq, DeepSeek, Ollama, Z.AI, OpenRouter — paste your key in the settings panel. No terminal auth dance. Multimodal (vision) models recommended — MudrikNow auto-screenshots every activation; Google Gemini Flash-Lite is a great free pick.
DemoSee it in 10 seconds
One hotkey. MudrikNow reads your screen, answers, and acts — no copy-pasting.
Hotkeys
Press a key, get help.
Reads your active window's full UI and anchors on what you're pointing at. MudrikNow opens on the opposite side of your screen, ready to help.
Quick chat — opens the panel instantly without capturing context. For questions that don't need screen awareness.
Cancel: stops streaming or closes the panel.
Send — Shift+Enter for a newline.
All hotkeys are rebindable from the ⚙ menu.
Privacy & security
Sandbox first.
Shell is read-only-only (mutation/piping blocked — violations kill the session); filesystem writes are off. UIA actions are allow-listed. A full-screen screenshot with a coordinate grid is captured on every Alt+Space (and via the Capture Context button) — the AI sees the pixels as well as the UI tree.
Read the full threat model in SECURITY.md →About
Hi, I'm Abdullah Magdy.
A dev who got tired of explaining context to AI chats — so I built MudrikNow on nights and weekends. It's open source; contributions and ideas are welcome 🤝
Let the owl see your screen.
Open-source, sandboxed, and one hotkey away. Download MudrikNow and stop explaining your screen to AI.
↓ Download