Everywhere is a context-aware AI assistant built for desktop users who want seamless, real-time AI assistance without switching apps or copying text. It captures screen content—whether it’s an error message, web article, or email draft—and delivers intelligent responses directly over your current application. Designed for developers, researchers, and professionals, it eliminates friction by bringing AI into your natural workflow.
Built with .NET 10 and Avalonia for native performance, Everywhere supports multiple LLM providers (OpenAI, Claude, Gemini, DeepSeek, Ollama), MCP tools for system integration, and works natively on Windows and macOS. It leverages RAG and UI automation to understand screen context, enabling powerful interactions like translation, summarization, and command execution—all triggered by a global keyboard shortcut.
What You Get
- Context-Aware Screen Capture - AI responds to text and UI elements visible on your screen without requiring copy-paste, using OCR and UI automation to understand context in real time.
- Multi-LLM Support - Unified access to OpenAI, Anthropic (Claude), Google (Gemini), DeepSeek, Moonshot (Kimi), SiliconCloud, and Ollama with seamless model switching.
- MCP Tool Integration - Direct integration with web browser, file system, terminal, and Windows Everything search to execute actions like file retrieval or command-line operations.
- Frosted Glass UI - Modern, translucent desktop interface that overlays any application with a sleek, non-intrusive chat interface for context-aware interaction.
- Global Keyboard Shortcuts - Instantly invoke the AI assistant from any app using customizable hotkeys, eliminating window switching and preserving workflow focus.
- Markdown-Rendered Responses - AI outputs are rendered with proper formatting, code blocks, and structure for readability and technical accuracy.
- Multi-Language Support - Full UI and AI responses in English, Chinese (Simplified/Traditional), German, Spanish, French, Italian, Japanese, Korean, Russian, and Turkish.
- System Command Execution - Use natural language to run shell commands, manage services, free ports, or clear caches with live terminal output and permission handling.
Common Use Cases
- Troubleshooting System Errors - A developer sees a cryptic error in their terminal; they press a hotkey, and Everywhere analyzes the text to suggest fixes and relevant commands.
- Quick Web Summarization - A researcher reads a 10-page technical paper and uses Everywhere to extract a 3-point summary without leaving the browser.
- Real-Time Translation of Foreign Text - A designer encounters a Japanese UI element and instantly translates it by selecting the text and invoking Everywhere.
- Professional Email Drafting - A manager writes a casual email and uses Everywhere to rewrite it in a polished, business-appropriate tone with context-aware suggestions.
Under The Hood
Architecture
- Clear separation of concerns through modular projects with dependency injection, enabling loose coupling between core logic and platform-specific components
- Plugin-based UI extensibility using polymorphic data templates and view-model abstractions in Avalonia, allowing dynamic rendering of diverse content types
- Async initialization pipelines with priority-based dependencies ensure non-blocking startup of critical systems like chat windows and hotkey listeners
- Build-time integration of a watchdog service as a pre-build publish target, ensuring native dependencies are compiled and deployed before the main application
Tech Stack
- .NET 10 with C# as the primary language, enforcing consistent naming and style through editorconfig and Roslyn-based analysis
- Strong emphasis on type safety via nullable reference types, ObservableObject patterns, and Avalonia’s reactive binding system
- No external databases or web frameworks detected, indicating a focused utility-layer or foundational codebase
- Windows-centric development environment with CRLF line endings and UTF-8-BOM encoding, suggesting targeted deployment scope
Code Quality
- Extensive test coverage with NUnit, including edge cases in encoding detection, caching, and object observation
- Robust error handling through defensive programming and null checks, prioritizing graceful degradation over exception throwing
- Comprehensive linting and test analysis tooling ensure code reliability without runtime assertions
- Innovative use of resilient reference and cache patterns to manage object lifetimes in memory-constrained scenarios
What Makes It Unique
- Native Avalonia integration for hardware-accelerated UI animations with dynamic delays based on visual position, eliminating external animation libraries
- Linux X11 window manipulation for non-interactive overlays with precise hit-test suppression and override-redirect control
- Auto-built watchdog process tightly coupled to the main application via build system, not external tooling
- Unified visual state management across platforms using Avalonia’s ElementComposition API, enabling code-defined animations with platform-agnostic semantics
- Custom screen selection UI with layered masks and adaptive tooltips that deliver native-like screen capture behavior without dependencies