# Kivixa

> A privacy-first cross-platform productivity workspace for notes, sketching, planning, and local AI assistance.

- **Type:** MCP server
- **Install:** `agentstack add mcp-990aa-kivixa`
- **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior
- **Seller:** [990aa](https://agentstack.voostack.com/s/990aa)
- **Installs:** 0
- **Category:** [Web & Browser](https://agentstack.voostack.com/c/web-and-browser)
- **Latest version:** 0.1.0
- **License:** MIT
- **Upstream author:** [990aa](https://github.com/990aa)
- **Source:** https://github.com/990aa/kivixa
- **Website:** https://kivixa.vercel.app/

## Install

```sh
agentstack add mcp-990aa-kivixa
```

Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent.

## About

# Kivixa

*Pronounced: kee-VEE-ha (/kiˈviːhɑː/)*

### A privacy-first cross-platform productivity workspace for notes, sketching, planning, and local AI assistance.

[](https://flutter.dev)
[](https://dart.dev)
[](LICENSE.md)
[](CHANGELOG.md)

**Install on Windows with winget (recommended):**

  winget install Kivixa

[](https://github.com/990aa/kivixa/releases/download/v0.9.0%2B90000/Kivixa-Setup-0.9.0.exe)

[](https://github.com/990aa/kivixa/releases/download/v0.9.0%2B90000/Kivixa-Android-0.9.0-arm64.apk)
[](https://github.com/990aa/kivixa/releases/download/v0.9.0%2B90000/Kivixa-Android-0.9.0-armv7.apk)
[](https://github.com/990aa/kivixa/releases/download/v0.9.0%2B90000/Kivixa-Android-0.9.0-x86_64.apk)

**F-Droid Repository (step-by-step):**

1. Install F-Droid from https://f-droid.org/.
2. Open the F-Droid app.
3. Go to **Settings**.
4. Tap **Repositories**.
5. Tap the **+** icon at the bottom.
6. Choose one method:
  - **Scan QR code** and scan the QR code below.
  - **Enter repository URL manually** and enter: `https://990aa.github.io/kivixa/repo/`

[](https://990aa.github.io/kivixa/repo/)

F-Droid repository QR code

---

## Features

### **On-Device AI**
Kivixa features a powerful on-device AI engine with multi-model support and Model Context Protocol (MCP) for intelligent, action-capable AI assistance without requiring an internet connection.

- **Multi-Model Support**
  - **Phi-4 Mini** - Default model for reasoning, conversation, and general assistance
  - **Phi-4 Mini Reasoning** - Reasoning-tuned Phi model for math-heavy and logic-heavy tasks
  - **Qwen 2.5 3B** - Specialized for writing, notes, and code generation
  - **Llama 3.2 3B Instruct** - Balanced compact model for MCP workflows, drafting, and general use
  - **Qwen2.5 1.5B Instruct** - Lightweight model for MCP actions and low-memory assistant workflows
  - **Qwen3.5 4B Distilled** - Claude 4.6 Opus reasoning-distilled 4B variant (Qwopus v3 GGUF source) for strongest Qwen quality
  - **Qwen3.5 2B Distilled** - Balanced speed and quality for daily work
  - **Qwen3.5 0.8B Distilled** - Fast lightweight option for low-memory devices
  - **DeepSeek R1 Distill Qwen 1.5B** - Compact reasoning model with strong math/code behavior
  - **SmolLM2 1.7B Instruct** - Fast compact instruct model for writing and lightweight coding
  - **SmolLM3 3B** - Newer compact SmolLM generation with stronger general assistant quality
  - **SmolVLM2 500M Video Instruct** - Compact vision-language model for image-aware prompting (GGUF + mmproj)
  - **Function Gemma 270M** - Ultra-fast, optimized for MCP tool calling (~180MB)
  - **Gemma 2B** - Google's efficient compact general-purpose model
  - **Gemma 3 4B IT** - Newer Gemma-family model with stronger balanced output quality
  - **Gemma 4 E2B IT** - Stronger Gemma-family instruct model for high-quality general, writing, and coding output
  - **TranslateGemma 4B IT** - Fast multilingual translation and rewriting model for notes
  - Seamless model switching for optimal performance

- **Model Context Protocol (MCP)**
  - **AI-Powered Tool Execution** - Let AI perform actions on your behalf
  - **File Operations** - Read, write, delete files; create folders
  - **Directory Listing** - Browse and explore your notes folder
  - **Markdown Export** - Export AI responses as formatted documents
  - **Lua Scripting Integration** - Execute calendar and timer scripts
  - **Safety First** - All tool executions require user confirmation
  - **Sandboxed Operations** - File access restricted to notes folder

- **Smart Model Manager**
  - Automatic download with resume support for AI models
  - Background downloading - continues even when app is minimized
  - Bundled asset downloads for multimodal models (model + mmproj companion files)
  - Progress tracking with speed and ETA display
  - GPU acceleration via Vulkan (Android/Windows/Linux) and Metal (macOS)

- **AI-Powered Features**
  - **Semantic Search** - Find notes by meaning, not just keywords
  - **Auto-Categorization** - Automatic topic extraction for organization
  - **Smart Summaries** - Generate concise summaries of long notes
  - **Question Answering** - Ask questions about your note content
  - **Title Suggestions** - AI-generated title recommendations
  - **MCP Chat Mode** - Toggle tool-enabled AI for file operations
  - **Prompt History Navigation** - Use Up/Down arrows in AI and MCP composers to cycle through your previous prompts (main chat and floating assistant)
  - **Attachment-Aware Composer** - Add multiple attachments with the `+` button, preview them above the composer, and remove any attachment before sending
  - **Attachment-Aware Responses** - On-device models receive attachment context (text extraction + binary metadata) for grounded responses
  - **Markdown-Aware Chat Rendering** - AI and MCP messages render markdown in chat while preserving raw markdown for model input
  - **Per-Response Model Metadata Export** - Exported chat JSON includes the model name/id that produced each assistant reply

- **Knowledge Graph Visualization**
  - **Interactive Mind Mapping** - Create visual knowledge networks
    - Pan and zoom navigation with touch/mouse gestures
    - Drag nodes to reposition them on the canvas
    - Grid background toggle for precise alignment
  - **Multiple Node Types**
    - Hub nodes for central topics
    - Note nodes that can link to your actual notes
    - Idea nodes for brainstorming
    - 6 shape options: Circle, Square, Diamond, Hexagon, Star, Rectangle
    - 8 color options with visual customization
  - **Rich Node Content**
    - Add titles and descriptions to each node
    - Link Note nodes to actual notes from your Browse section
    - View and open linked notes directly from the graph
    - Supports handwritten (.kvx), markdown (.md), and text files
  - **Flexible Link System**
    - Connect nodes with customizable links
    - Add text labels to links (e.g., "relates to", "depends on")
    - 3 line thickness options: Thin, Normal, Thick
    - Arrow styles: None, Single Arrow, Double Arrow
    - Custom link colors
    - Manage links dialog to view/delete connections
  - **Navigation & Organization**
    - Recenter view to all nodes
    - Focus on selected node
    - Clear all nodes/links or load demo data
    - Persistent storage - your graph is saved automatically

- **Vector Database**
  - Local vector embeddings for all your notes
  - Lightning-fast similarity search
  - Automatic clustering of related content
  - Persistent cache for instant startup

- **Privacy-First AI**
  - Runs 100% on-device - no data leaves your device
  - No internet required after initial model download
  - No API keys or subscriptions needed
  - Your notes stay private

- **Audio Intelligence**
  - **Speech-to-Text (STT)** - Whisper-based offline transcription
    - Multiple model sizes (Tiny to Large) for accuracy/speed tradeoffs
    - Real-time streaming transcription with word-level timestamps
    - Semantic audio indexing - search by spoken content
    - Multi-language support with automatic detection
  - **Text-to-Speech (TTS)** - Kokoro neural speech synthesis
    - Natural-sounding voices (Male, Female, Custom)
    - Adjustable rate and pitch controls
    - Word boundary tracking for lip-sync
    - High-quality 24kHz audio output
  - **Voice Activity Detection (VAD)**
    - Efficient speech/silence detection
    - Automatic noise floor calibration
    - Saves battery by only processing speech
  - **Neural Dictation Bar** - Smart voice input accessory
    - Text and Command dictation modes
    - Live confidence indicator
    - Seamless keyboard integration
    - Auto-insert into text fields
  - **Voice Notes** - Record and transcribe voice memos
    - Interactive waveform visualization (bars, line, circular, orb)
    - Karaoke-style transcript highlighting
    - Search within transcript
    - Speaker identification support
  - **Voice Search** - Search notes by voice
    - Press-and-hold or tap-to-toggle modes
    - Real-time transcription preview
    - Confidence-based result filtering
  - **AI Walkie-Talkie** - Hands-free AI conversations
    - Full-screen immersive interface
    - Dual animated orbs (user/AI)
    - Conversation history
    - Push-to-talk or auto-listen modes
  - **Read Aloud** - TTS accessibility features
    - Mini player with playback controls
    - Speed adjustment (0.5x - 2.0x)
    - Sentence-by-sentence navigation
    - Voice selection
    - Floating action button integration
  - **Fully Offline** - All audio processing runs locally

### **Notes & Documents**
- **Rich Markdown Editor** - Create beautiful formatted documents with AppFlowy Editor
  - Text formatting (bold, italic, underline, strikethrough)
  - Block formatting (lists, checkboxes, quotes, code blocks)
  - Hyperlink support
  - Real-time autosave
- **Floating Text Boxes** - Add moveable, resizable text boxes anywhere on the canvas
- **File Organization** - Intuitive folder system with move, rename, and delete operations
- **Multiple Formats** - Support for `.kvx` (native), `.md` (markdown), `.txt` (text), and PDF files
- **Text File Editor** - Full-featured text editor with syntax highlighting
- **Note Linking** - Connect related notes together with bidirectional links

### **Media Embedding**
- **Image & Video Upload** - Embed media directly into markdown and text files
  - Upload from local storage or paste from web URLs
  - Insert local absolute paths (``)
  - Automatic file management with dedicated media storage
  - Videos are fully playable within embedded player with volume control
- **Interactive Media Controls**
  - **Resize**: Drag corner/edge handles to adjust dimensions
  - **Rotate**: Rotate images with 15° snapping for precision
  - **Move Handle**: 4-way arrow in center for precise repositioning
  - **Drag**: Reposition media anywhere within the document
  - **Aspect Ratio Lock**: Hold Shift while resizing to maintain proportions
  - **Keyboard Support**: Escape to deselect, Shift for aspect lock
- **Comment Annotations**
  - Add optional comments to any image or video
  - Hover to reveal (Windows) or tap icon (Android)
  - Edit and delete comments easily
  - URL-encoded for special character support
- **Web Image Modes** (configurable in Settings)
  - **Download Locally**: Cache images for offline access
  - **Fetch on Demand**: Load from web each time (saves storage)
  - Cache management with size display and clear option
- **Large Image Preview**
  - Scrollable preview for images > 2000px
  - Pan and zoom within a constrained container
  - Minimap showing visible region
  - Toggle between preview and full display modes
  - Configurable preview container size (100-500px)
- **Extended Markdown Syntax**
  - ``
  - All transforms persist in standard markdown files
  - Compatible with other markdown viewers (shows default)
- **Performance Optimized**
  - LRU caching for loaded images
  - Thumbnail generation for heavy media
  - Lazy loading via visibility detection
  - Isolated repaints with RepaintBoundary
  - Auto-hide video controls during playback

### **Life Git (Version Control)**
- **Time Travel** - Roll back any note to any previous version with an intuitive slider
- **Auto-Snapshots** - Automatic versioning when you stop typing (2-second debounce)
- **Content-Addressable Storage** - Git-like blob storage using SHA-256 hashing
- **Commit History** - Browse all changes with timestamps and commit messages
- **Version Comparison** - Preview any historical version before restoring
- **Per-File History** - Track changes for individual files
- **Zero Configuration** - Just write, Life Git handles versioning automatically

### **Scriptable Plugin System**
- **Lua Scripting** - Automate tasks with Lua 5.3 scripts
- **Built-in App API** - Access your notes programmatically:
  - `App:createNote(path)` - Create new notes
  - `App:readNote(path)` - Read note content
  - `App:writeNote(path, content)` - Write to notes
  - `App:getAllNotes()` - List all notes
  - `App:findNotes(query)` - Search notes
  - `App:getRecentNotes(count)` - Get recent notes
  - `App:moveNote(from, to)` - Move notes
  - `App:deleteNote(path)` - Delete notes
  - `App:getStats()` - Get workspace statistics
- **Example Plugins Included** - Archive tasks, daily summaries, move overdue items
- **Plugin Manager** - Enable/disable and run plugins from the UI
- **Script Runner** - Execute ad-hoc Lua code directly

### **Project Manager**
- **Project Dashboard** - Organize your work into dedicated projects
- **Task Management** - Create and track tasks within projects
- **Progress Tracking** - Visual indicators for project completion status
- **Project Categories** - Organize projects with custom categories

### **Calendar & Events**
- **Interactive Calendar** - View and manage events with day, week, and month views
- **Event Management** - Create, edit, and delete events with custom colors
- **Recurring Events** - Support for daily, weekly, monthly, and yearly recurrence
- **Reminders** - Get notified before important events
- **Calendar Navigation** - Easy navigation between dates with jump-to-date feature

### **Digital Canvas**
- **Advanced Drawing Tools**
  - Pen with pressure sensitivity
  - Highlighter for emphasis
  - Laser pointer for presentations
  - Eraser with multiple modes
  - Shape tools and selection
- **Professional Canvas Features**
  - Infinite canvas with smooth pan & zoom
  - Layer support for complex artwork
  - Background customization (colors, patterns, images)
  - Grid and snap-to-grid options
  - Transform tools (rotate, scale, move)

### **PDF Integration**
- Import and annotate PDF documents
- Export notes as PDFs
- Maintain annotations across devices
- High-quality rendering

### **Productivity**
- **Smart Autosave** - Never lose your work with automatic saving
- **Split Screen View** - Open and edit two files simultaneously side-by-side
  - Support for different file types (handwritten, markdown, text)
  - Horizontal and vertical split modes
  - Resizable panes with drag-to-adjust divider
  - Swap panes functionality
- **File Manager** - Powerful organization with search and filtering
- **Recent Files** - Quick access to your latest work
- **Cross-Platform Sync** - Work seamlessly across devices
- **Keyboard Shortcuts** - Boost productivity with keybindings

### **Productivity Clock**
A comprehensive productivity timer system with advanced features for focused work sessions.

- **Core Timer Features**
  - **Pomodoro Technique** - Classic 25-minute focus sessions with breaks
  - **Multiple Timer Templates** - Pomodoro (25/5), 52/17 Method, Ultradian (90/20)
  - **Custom Durations** - Set any work/break duration you prefer
  - **Session Types** - Focus, Short Break, Long Break, Flow, and Custom modes
  - **Progress Tracking** - Visual circular progress indicator
  - **Floating Clock Widget** - Always-visible timer overlay while working

- **Context-Aware Timers**
  - **Session Tags** - Tag sessions with context (Coding, Reading, Writing, Meeting, Research, Design, Learning, Planning, Exercise)
  - **Filter Stats by Tag** - Analyze productivity by activity type
  - **Custom Tags** - Create your own context tags with custom colors and icons

- **Quick-Switch Presets**
  - **Code Preset** - 90-minute deep work with 20-minute breaks
  - **Reading Preset** - 45-minute reading with 10-minute breaks
  - **Deep Design Preset** - 2-hour creative sessions with 25-minute breaks
  - **Sprint Preset** - Quick 25-minute bursts with 5-minute breaks
  - **Meetings Preset** - 30-minute time-boxed meetings with 5-minute prep

- **Multi-Timer Orchestration**
  - **Parallel Clocks** - Run main focus timer with secondary reminders
  - **Built-in Presets** - Tea timer, Commit reminder, Eye rest (20-20-20), Stretch, Water break, Standup, Posture check, Meeting alert
  - **Custom Timers** - Create unlimited secondary countdown timers
  - **Independent Control** - St

…

## Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [990aa](https://github.com/990aa)
- **Source:** [990aa/kivixa](https://github.com/990aa/kivixa)
- **License:** MIT
- **Homepage:** https://kivixa.vercel.app/

Install and usage instructions live in the source repository linked above.

## Pricing

- **Free** — Free

## Security capabilities

Automated source analysis of v0.1.0 — what this tool can access:

- **Network access:** no
- **Filesystem access:** no
- **Shell / process execution:** no
- **Environment & secrets:** no
- **Dynamic code execution:** no

*"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.*


## Versions

- **0.1.0** — security scan: passed — Imported from the upstream source.

## Links

- Listing page: https://agentstack.voostack.com/l/mcp-990aa-kivixa
- Seller: https://agentstack.voostack.com/s/990aa
- Browse the marketplace: https://agentstack.voostack.com/browse

---
Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.
