diff --git a/CLAUDE.md b/CLAUDE.md
index 05bfb5254..8bdd7c235 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -25,23 +25,54 @@ Set `DOCKER_ENABLE_SECURITY=true` environment variable to enable security featur
 - **Proxy Configuration**: Vite proxies `/api/*` calls to backend (localhost:8080)
 - **Build Process**: DO NOT run build scripts manually - builds are handled by CI/CD pipelines
 - **Package Installation**: DO NOT run npm install commands - package management handled separately
+- **Deployment Options**:
+  - **Desktop App**: `npm run tauri-build` (native desktop application)
+  - **Web Server**: `npm run build` then serve dist/ folder
+  - **Development**: `npm run tauri-dev` for desktop dev mode
 
-#### Tailwind CSS Setup (if not already installed)
-```bash
-cd frontend
-npm install -D tailwindcss postcss autoprefixer
-npx tailwindcss init -p
-```
+#### Multi-Tool Workflow Architecture
+Frontend designed for **stateful document processing**:
+- Users upload PDFs once, then chain tools (split → merge → compress → view)
+- File state and processing results persist across tool switches
+- No file reloading between tools - performance critical for large PDFs (up to 100GB+)
+
+#### FileContext - Central State Management
+**Location**: `src/contexts/FileContext.tsx`
+- **Active files**: Currently loaded PDFs and their variants
+- **Tool navigation**: Current mode (viewer/pageEditor/fileEditor/toolName)  
+- **Memory management**: PDF document cleanup, blob URL lifecycle, Web Worker management
+- **IndexedDB persistence**: File storage with thumbnail caching
+- **Preview system**: Tools can preview results (e.g., Split → Viewer → back to Split) without context pollution
+
+**Critical**: All file operations go through FileContext. Don't bypass with direct file handling.
+
+#### Processing Services
+- **enhancedPDFProcessingService**: Background PDF parsing and manipulation
+- **thumbnailGenerationService**: Web Worker-based with main-thread fallback
+- **fileStorage**: IndexedDB with LRU cache management
+
+#### Memory Management Strategy
+**Why manual cleanup exists**: Large PDFs (up to 100GB+) through multiple tools accumulate:
+- PDF.js documents that need explicit .destroy() calls
+- Blob URLs from tool outputs that need revocation
+- Web Workers that need termination
+Without cleanup: browser crashes with memory leaks.
+
+#### Tool Development
+- **Pattern**: Follow `src/tools/Split.tsx` as reference implementation
+- **File Access**: Tools receive `selectedFiles` prop (computed from activeFiles based on user selection)
+- **File Selection**: Users select files in FileEditor (tool mode) → stored as IDs → computed to File objects for tools
+- **Integration**: All files are part of FileContext ecosystem - automatic memory management and operation tracking
+- **Parameters**: Tool parameter handling patterns still being standardized
+- **Preview Integration**: Tools can implement preview functionality (see Split tool's thumbnail preview)
 
 ## Architecture Overview
 
 ### Project Structure
 - **Backend**: Spring Boot application with Thymeleaf templating
-- **Frontend**: React-based SPA in `/frontend` directory (replacing legacy Thymeleaf templates)
-  - **Current Status**: Active development to replace Thymeleaf UI with modern React SPA
+- **Frontend**: React-based SPA in `/frontend` directory (Thymeleaf templates fully replaced)
   - **File Storage**: IndexedDB for client-side file persistence and thumbnails
   - **Internationalization**: JSON-based translations (converted from backend .properties)
-  - **URL Parameters**: Deep linking support for tool states and configurations
 - **PDF Processing**: PDFBox for core PDF operations, LibreOffice for conversions, PDF.js for client-side rendering
 - **Security**: Spring Security with optional authentication (controlled by `DOCKER_ENABLE_SECURITY`)
 - **Configuration**: YAML-based configuration with environment variable overrides
@@ -59,9 +90,8 @@ npx tailwindcss init -p
 - **Pipeline System**: Automated PDF processing workflows via `PipelineController`
 - **Security Layer**: Authentication, authorization, and user management (when enabled)
 
-### Template System (Legacy + Modern)
-- **Legacy Thymeleaf Templates**: Located in `src/main/resources/templates/` (being phased out)
-- **Modern React Components**: Located in `frontend/src/components/` and `frontend/src/tools/`
+### Component Architecture
+- **React Components**: Located in `frontend/src/components/` and `frontend/src/tools/`
 - **Static Assets**: CSS, JS, and resources in `src/main/resources/static/` (legacy) + `frontend/public/` (modern)
 - **Internationalization**:
   - Backend: `messages_*.properties` files
@@ -91,13 +121,14 @@ npx tailwindcss init -p
    - Frontend: Update JSON files in `frontend/public/locales/` or use conversion script
 5. **Documentation**: API docs auto-generated and available at `/swagger-ui/index.html`
 
-## Frontend Migration Notes
+## Frontend Architecture Status
 
-- **Current Branch**: `feature/react-overhaul` - Active React SPA development
-- **Migration Status**: Core tools (Split, Merge, Compress) converted to React with URL parameter support
-- **File Management**: Implemented IndexedDB storage with thumbnail generation using PDF.js
-- **Tools Architecture**: Each tool receives `params` and `updateParams` for URL state synchronization
-- **Remaining Work**: Convert remaining Thymeleaf templates to React components
+- **Core Status**: React SPA architecture complete with multi-tool workflow support
+- **State Management**: FileContext handles all file operations and tool navigation  
+- **File Processing**: Production-ready with memory management for large PDF workflows (up to 100GB+)
+- **Tool Integration**: Standardized tool interface - see `src/tools/Split.tsx` as reference
+- **Preview System**: Tool results can be previewed without polluting file context (Split tool example)
+- **Performance**: Web Worker thumbnails, IndexedDB persistence, background processing
 
 ## Important Notes
 
@@ -108,6 +139,11 @@ npx tailwindcss init -p
   - **Backend**: Designed to be stateless - files are processed in memory/temp locations only
   - **Frontend**: Uses IndexedDB for client-side file storage and caching (with thumbnails)
 - **Security**: When `DOCKER_ENABLE_SECURITY=false`, security-related classes are excluded from compilation
+- **FileContext**: All file operations MUST go through FileContext - never bypass with direct File handling
+- **Memory Management**: Manual cleanup required for PDF.js documents and blob URLs - don't remove cleanup code
+- **Tool Development**: New tools should follow Split tool pattern (`src/tools/Split.tsx`)
+- **Performance Target**: Must handle PDFs up to 100GB+ without browser crashes
+- **Preview System**: Tools can preview results without polluting main file context (see Split tool implementation)
 
 ## Communication Style
 - Be direct and to the point