Anarlog — Open-Source AI Meeting Notes That Stay on Your Machine

Introduction

Anarlog is an open-source alternative to Granola and similar AI meeting assistants. It captures meeting audio locally, transcribes it using on-device models, and generates structured summaries. All data stays on your machine, making it suitable for confidential meetings and organizations with strict data policies.

What Anarlog Does

• Captures system and microphone audio during meetings
• Transcribes speech to text using local AI models
• Generates structured meeting summaries with action items
• Organizes notes by meeting with searchable history
• Exports notes in Markdown and other formats

Architecture Overview

Anarlog is built with Tauri v2 (Rust backend, web frontend) for a small binary size and native performance. Audio capture uses platform APIs, and transcription runs locally via Whisper-derived models. Summarization can use local LLMs or optionally connect to an API. The frontend is a React-based note editor that displays real-time transcription and AI-generated summaries side by side.

Self-Hosting & Configuration

• Download prebuilt binaries from GitHub Releases
• No account creation or API keys required for basic operation
• Optionally configure an LLM API for enhanced summarization
• All meeting data stored locally in application data directory
• Supports macOS, Windows, and Linux

Key Features

• Local-first architecture with no mandatory cloud dependency
• Real-time transcription during meetings
• AI-powered summary generation with action items
• Clean note-taking interface with meeting timeline
• Built with Tauri for small footprint and fast startup

Comparison with Similar Tools

• Granola — closed source and cloud-dependent; Anarlog is open source and local
• Otter.ai — cloud transcription with subscription; Anarlog is free and offline
• Fireflies.ai — records via meeting bot; Anarlog captures audio locally without joining calls
• tl;dv — integrates with Zoom/Meet; Anarlog works with any audio source
• Screenpipe — continuous full-screen recording; Anarlog focuses specifically on meeting workflows

FAQ

Q: Does it join my video call as a bot? A: No. Anarlog captures audio from your system output and microphone directly, so no bot joins the meeting.

Q: Can I use it without any AI API? A: Yes. Local transcription works without any API. Summarization can use a local LLM via Ollama.

Q: What meeting platforms does it work with? A: Any platform that produces audio on your machine: Zoom, Google Meet, Teams, Discord, or in-person meetings via microphone.

Q: How accurate is the local transcription? A: Quality depends on the Whisper model size selected. The large model produces near-professional accuracy for English.

Sources

• https://github.com/fastrepl/anarlog