Cursivis-Next

A living cursor that understands context.

Comment

Cursivis Next

The most frustrating part of using AI isn't getting the answer. It's stopping your work to explain everything the AI could already see.

💡 Inspiration

The usual AI workflow breaks concentration:

  1. Copy content
  2. Leave your application
  3. Open a chatbot
  4. Paste and reconstruct context
  5. Wait for response
  6. Copy answer back

Real work doesn't begin in a blank prompt box. It begins in the browser tab, code editor, document, or email already open on your screen.

The Core Question

What if AI could understand the context already in front of you?

Cursivis Next answers this with a simple interaction model:

Selection = Context
Trigger = Intent
Cursivis = Action

We didn't build another chatbot. We built an assistant that appears when needed, helps you move forward, and disappears back into your workflow.

✨ What It Does

Cursivis Next is a context-aware Windows AI assistant that works across all your applications.

Core Workflow

Select → Trigger → Understand → Respond → Refine or Act

Highlight text, press a global hotkey, and Cursivis adapts:

Context Result
Article Concise summary
Difficult paragraph Simplified explanation
Foreign text Translation
Email draft Professional rewrite
Code snippet Explanation & review
Messy notes Structured action items
Rough prompt Polished AI instruction
Image/screen region Visual analysis

One interaction. Infinite adaptations.

Key Features

🎯 Smart Mode Instantly infers the best transformation and executes it—no menus, no choices, just results.

🧭 Guided Mode Dynamically generates contextual options (Summarize, Explain, Rewrite, Translate, Extract) based on your actual selection.

👁️ Screen & Image Understanding Select any screen region when text isn't available. Analyzes screenshots, error messages, diagrams, and visual documents.

💬 Live Mode Real-time voice conversation with bounded tools:

  • Refresh selected text
  • Analyze screen regions
  • Insert text into applications
  • Navigate complex interfaces
  • User-controlled memory

🎤 Smart Dictation Transcribes speech, improves punctuation and structure, inserts polished text directly where you're working.

🔧 Prompt Optimizer Transforms incomplete instructions into clear, structured prompts before sending to any AI system.

🧭 Navigation Guidance Step-by-step task assistance: observe → instruct → wait → verify → continue. You stay in control.

⚡ Take Action Structured browser automation with safety validation—fills forms, handles interactions, requires confirmation for sensitive operations.

🏗️ How We Built It

Technology Stack

  • Platform: C# + .NET 8 + WinUI 3
  • AI Integration: OpenAI Responses, Realtime, & Transcription APIs
  • Windows APIs: UI Automation, Global Hotkeys, Audio Capture
  • Browser: Chromium Extension + Native Messaging
  • Installer: Inno Setup

Architecture

Global Keyboard Shortcuts
        ↓
Windows Context Capture
        ↓
Application Layer
        ↓
AI Services (Responses, Realtime, Audio)
        ↓
Result / Voice / Guidance / Action

Technical Highlights

Context Capture Multi-strategy approach combining:

  • Windows UI Automation
  • Protected clipboard fallback
  • Foreground window validation
  • Clipboard restoration
  • Cross-app compatibility (browsers, editors, documents, PDFs)

Desktop Overlays Native Windows integration:

  • Transparent compositor backdrops
  • DPI-aware positioning
  • Draggable/resizable surfaces
  • Persistent placement
  • Non-activating windows

Browser Action Layer

  • Chromium extension
  • Authenticated local bridge
  • Structured action planning
  • Safety classification
  • Confirmation for sensitive operations

Privacy & Security

  • Local settings storage
  • Windows encrypted credential storage
  • No ambient monitoring
  • Explicit user consent for screen capture

🚧 Challenges We Ran Into

1. Reliable Selection Capture

Different apps expose text differently. Chrome selected "extempore" but returned the browser URL instead.

Solution: Multi-fallback strategy with browser-specific safeguards and real cross-application testing.

2. Global Shortcuts

Shortcuts owned by other apps, thread ownership issues, registration rollback failures.

Solution: Transactional registration on WinUI owner thread with conflict detection, visible fallback reporting, and persistence of configured vs. active shortcuts.

3. Transparent Overlays

Opaque swap chains, black edges, DPI-dependent clipping, focus stealing.

Solution: Native window regions, compositor handling, DPI-aware geometry, screenshot-based validation.

4. Realtime Interaction

Coordinating microphone capture, audio buffering, interruption, tool calls, transcripts, and cancellation.

Solution: Managed session lifecycle with coherent state transitions across all interaction points.

5. Balancing Action with Safety

Enabling useful actions without creating risk.

Solution: Bounded tools, structured validation, explicit permission, confirmation UI, and undo support.

🏆 Accomplishments We're Proud Of

✅ Shortcut-driven AI assistant that works system-wide
✅ Non-destructive clipboard capture with restoration
✅ Smart Mode + Guided Mode adaptive intelligence
✅ Image-region analysis when text is unavailable
✅ Context preservation across refinement, insertion, replacement, undo
✅ Voice mode with tools (screen analysis, text insertion, navigation)
✅ Explicit user-controlled memory (no silent conversation saving)
✅ Navigation Guidance that observes, waits, and verifies
✅ Structured browser automation with safety boundaries
✅ Visible shortcut conflict resolution
✅ Clean installation isolated from previous versions
✅ Windows encrypted API key storage
✅ Self-contained installer with startup integration

Quality Metrics

  • 167 unit tests ✅
  • 71 integration tests ✅
  • Clean x64 Release build (zero warnings, zero errors) ✅

Tests cover invisible reliability:

  • Clipboard restoration
  • Shortcut persistence & rollback
  • Thread ownership
  • Cancellation
  • Storage migration
  • Safety validation
  • Error recovery

📚 What We Learned

The Biggest Lesson

The quality of an AI product is determined by far more than the intelligence of its model.

A powerful model cannot compensate for:

  • Capturing the wrong selection
  • Losing a keyboard shortcut
  • Stealing focus
  • Damaging clipboard contents
  • Displaying broken overlays
  • Forgetting context
  • Executing unintended actions

Key Insights

  • Context capture is often harder than reasoning
  • Selection is a better prompt than a blank text box
  • A shortcut expresses intent faster than another menu
  • Invisible reliability is a real product feature
  • Users trust actions when they're visible, bounded, reversible, and cancellable
  • Desktop AI must respect focus, DPI, accessibility, clipboard state, OS lifecycle, and native thread ownership

Most importantly: AI doesn't need to dominate the workflow to transform it. It can appear for a few seconds, understand what matters, help you move forward, and disappear again.

🚀 What's Next

Immediate Goals

  • Polish into a dependable Windows productivity layer
  • Improve selection reliability across all application types
  • Enhance Realtime voice stability and interruption handling
  • Strengthen image and screen understanding
  • Refine Navigation Guidance
  • Expand safe browser workflows
  • Improve accessibility and keyboard-first interaction

Long-Term Vision

  • Developer capability system for extensible context tools
  • macOS version using AppKit
  • Local model integration (Ollama, LM Studio)
  • Application-specific context adapters
  • Evaluation systems for action correctness and safety
  • Seamless installer and update experience

The Ultimate Goal

Cursivis should make AI feel less like a destination and more like a capability already present inside the work.

The cursor should not merely point at information.
The selection should not merely highlight it.
The shortcut should not merely open another application.

Together, they should become the beginning of understanding—and, when you choose, the beginning of action.

🔗 Links

Built With

  • accessibility
  • browser-automation
  • chromium-extension
  • clipboard
  • computer-vision
  • context-aware
  • csharp
  • desktop-app
  • dotnet
  • hotkeys
  • image-analysis
  • inno-setup
  • multimodal
  • native-messaging
  • navigation-guidance
  • openai
  • productivity
  • realtime-api
  • responses-api
  • screen-capture
  • smart-dictation
  • ui-automation
  • voice-ai
  • windows
Share this project:

Updates

Submission history