Natively offers an open-source, privacy-first AI assistant for real-time support during meetings and conversations. It functions as an undetectable, always-on-top desktop overlay, providing instant, context-aware assistance without recording. This tool targets professionals and students needing immediate, intelligent input during live interactions, enhancing productivity while maintaining strict data control.

How It Works

Natively captures audio and screen content locally via an invisible desktop overlay. It maintains a rolling context window for ongoing conversations and processes data locally. Prompt data is sent to user-configured AI providers (cloud or local Ollama), prioritizing low latency and real-time responsiveness with minimal data transmission.

Quick Start & Requirements

End users can download prebuilt Windows (v1.1.2) and macOS (v1.1.3) versions from Releases, requiring no build steps. Developers require Node.js (v20+ recommended), Git, and Rust. Essential prerequisites include a Google Gemini API Key and a Google Cloud Service Account for real-time speech-to-text (a hard dependency). Setup involves cloning the repository, running npm install, configuring API keys and the Google service account JSON path in a .env file, and then running npm start for development or npm run dist for production builds.

Highlighted Details

• Invisible, always-on-top desktop overlay for seamless integration.
• Real-time transcription and rolling context memory across conversations.
• Screenshot and screen content analysis for immediate understanding of visual information.
• Supports multiple AI providers including Gemini, OpenAI, Claude, Groq, and local Ollama.
• Privacy-centric design with local data storage, Bring Your Own Keys (BYOK), and no telemetry.

Maintenance & Community

The project is marked as active and welcomes contributions for bug fixes, feature improvements, and documentation. It is open to sponsorships, ads, or partnerships. Specific community channels like Discord or Slack are not detailed in the README.

Licensing & Compatibility

Natively is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). This strong copyleft license requires that any derivative work or software run over a network must also be made available under the AGPL-3.0. This may impose significant restrictions on integration with closed-source commercial products.

Limitations & Caveats

Linux support is currently limited and actively seeking maintainers. Real-time transcription is critically dependent on a configured Google Cloud Speech-to-Text service account, which requires setup and potential associated costs with Google Cloud services.