# Lumina > Lumina is a free, source-available, local-first meeting notes and knowledge base app for macOS, Windows and Linux, and an open alternative to Google AI Edge Foresight. It records the microphone and the computer's audio, transcribes live on the CPU, turns the user's shorthand into enhanced notes aligned line by line (three views: side by side like Foresight, Cornell rows, or one at a time), writes a summary (overview, decisions, action items, open questions), answers questions asked in meetings from the user's documents or, when they do not have it, the web (live assistance), indexes files, folders and web pages (their text or an AI summary), lays out summaries in five styles or a custom one, offers a chat with citations over everything (it reads attached files and audio, and creates, edits or deletes notes when asked), and exposes the knowledge base to AI agents through an MCP server. No account, no telemetry; everything runs on the device. Key facts: - Install on macOS/Linux: `curl -fsSL https://lautaro005.github.io/lumina/install.sh | sh`. On Windows (PowerShell): `irm https://lautaro005.github.io/lumina/install.ps1 | iex`. Start with `lumina` (browser) or `lumina app` (desktop app with tray icon). - Current version: 0.2.1 (November 1, 2026). - License: Apache-2.0 with the Commons Clause: free to use, modify and share, not to sell. - Speech models: Parakeet TDT 0.6B v3 (25 European languages, auto-detects), Canary 1B v2 (same 25 languages, follows the chosen language), Whisper tiny to large-v3-turbo (99 languages), Silero VAD, plus an experimental English-only option built on EmbeddingGemma 2's audio encoder. The transcription language is a setting with "Automatic" plus every supported language. - Embeddings: EmbeddingGemma 2 (8-bit ONNX, 768 dimensions, truncatable to 512/256/128), downloaded from the app. Vector store: ChromaDB (collections documents, meetings, notes). - Notes model: Gemma 4 E2B, E4B or 12B (4-bit GGUF) run by llama.cpp inside Lumina, or Ollama, LM Studio or any OpenAI-compatible server. - Optional read-aloud: Kokoro 82M or the system voice. Optional web search in chat: DuckDuckGo or SearXNG, off by default. Live assistance searches the web (the question only) when the sources do not answer; switchable. - MCP tools: search_knowledge, list_sources, get_document, add_note, add_web_page, query_meetings, get_meeting, rewrite_summary, save_meeting_to_brain. - Connected apps (0.1.7): saves meetings to brain and searches brain in its chat; sends meetings to OpenCue agents; embedded by AI WorkStudio, whose MCP hub includes Lumina's tools. stdio: `lumina mcp` (e.g. `claude mcp add lumina -- lumina mcp`). HTTP: `http://127.0.0.1:47770/mcp` with a bearer token from Settings → Agents. - Stack: Python (FastAPI) server on 127.0.0.1:47770, React + TypeScript interface, Tauri 2 desktop shell. Data folder: `~/.lumina/data`. - Updates: Lumina checks GitHub for a newer version (switchable) and updates itself from a button in the app. Models not in use are unloaded after a few minutes ("Free memory when idle"). - System audio: ScreenCaptureKit on macOS, WASAPI loopback on Windows, PulseAudio/PipeWire monitor on Linux. - Interface languages: English and Spanish. ## Docs - [Documentation](https://lautaro005.github.io/lumina/docs/): install, first run, meetings, languages and transcription, notes model, sources, chat, read aloud, MCP, data, commands, troubleshooting, build from source - [Install](https://lautaro005.github.io/lumina/docs/#install): one-line installers and options - [Languages](https://lautaro005.github.io/lumina/docs/#languages): which speech model handles which language - [Agents (MCP)](https://lautaro005.github.io/lumina/docs/#agents): tools and client configuration ## Project - [Changelog](https://lautaro005.github.io/lumina/changelog/): what changed in each version - [Privacy policy](https://lautaro005.github.io/lumina/privacy/): what stays on the device and when the network is used - [Security policy](https://lautaro005.github.io/lumina/security/): how to report vulnerabilities privately - [Source code](https://github.com/Lautaro005/lumina): repository, README and issues ## Optional - [Releases](https://github.com/Lautaro005/lumina/releases): desktop installers for every version