Repolode
Submit
marvis-ai

marvis-ai/marvis

A privacy-first desktop AI co-pilot that sees your screen on request, listens to meetings, and keeps keys and data local.

Open-source alternative to:Otter.ai

View on GitHubWebsite
Stars
54
Forks
5
Open issues
1
Last commit
11 days ago

What is it?

Marvis is a desktop assistant that floats above your workspace as a small translucent bar. You can ask it questions about what is on your screen, and it sends the latest screen frame along with your prompt to the LLM you configure, streaming the answer back as markdown. Screen frames are kept only in an in-memory ring buffer and are never written to disk.

The Listen mode captures microphone and system audio as separate channels, transcribes them with either Deepgram or a local whisper.cpp model, and produces a live transcript with speaker labels plus rolling summaries. Providers include OpenAI, Anthropic, Gemini, OpenRouter, Ollama, and any OpenAI-compatible endpoint, arranged as a failover chain.

The app is built with Tauri 2, with the Rust core handling capture, audio, keys, and storage, and a React webview for the UI. Sessions and transcripts are stored in a local SQLite database under ~/.marvis. It supports macOS, Windows, and Linux.

Key features

  • 01Frameless overlay bar that docks to any screen edge and stays on top
  • 02Ask about the current screen with a global hotkey, with frames kept only in memory
  • 03Dual-channel meeting capture with live transcript and rolling AI summaries
  • 04Bring your own model, including local Ollama and any OpenAI-compatible endpoint
  • 05Provider failover chain that hands off to the next provider when a call fails
  • 06Local SQLite storage for sessions, transcripts, and summaries

Use cases

  • Asking an LLM for help with whatever is currently on your screen without uploading screenshots to storage
  • Transcribing and summarizing meetings locally with whisper.cpp
  • Running a desktop AI assistant against a local Ollama model with no cloud account
  • Reviewing and exporting past meeting transcripts from a local database

Quick start

  1. 1

    Install dependencies with Bun

    bun install
  2. 2

    Run the desktop app in development mode (requires Rust toolchain and Bun 1.3+)

    bun run build:dev

Things to know

  • Requires a Rust stable toolchain and platform-specific build tools to run from source.
  • Needs at least one LLM provider: an API key or a running Ollama daemon.
  • API keys are stored in plaintext in ~/.marvis/keys.json with no encryption.
  • Screen capture and system audio behave differently per OS, and Linux needs extra system packages.

In collections

Open-source AI agent frameworks & appsRun LLMs locallyPrivacy & local-first toolsAI video & audio tools

Topics

#local-ai#screen-capture#speech-to-text#meeting-notes#tauri#rust#privacy#llm#ai#ai-agents

FAQ

What is marvis?
A privacy-first desktop AI co-pilot that sees your screen on request, listens to meetings, and keeps keys and data local. Marvis is a desktop assistant that floats above your workspace as a small translucent bar. You can ask it questions about what is on your screen, and it sends the latest screen frame along with your prompt to the LLM you configure, streaming the answer back as markdown. Screen frames are kept only in an in-memory ring buffer and are never written to disk.
Is marvis free to use? What license does it have?
marvis is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is marvis still maintained?
Actively maintained. The last commit to marvis was on Oct 12, 2026.
What can I use marvis for?
Asking an LLM for help with whatever is currently on your screen without uploading screenshots to storage Transcribing and summarizing meetings locally with whisper.cpp Running a desktop AI assistant against a local Ollama model with no cloud account Reviewing and exporting past meeting transcripts from a local database
How do I install marvis?
Install dependencies with Bun: bun install Run the desktop app in development mode (requires Rust toolchain and Bun 1.3+): bun run build:dev
Where can I get marvis?
The source code and installation instructions are on GitHub: https://github.com/marvis-ai/marvis

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects