Anarlog: An Open-Source, Local-First Granola Alternative That Takes Meeting Notes Without a Bot

By Prahlad Menon 3 min read

AI meeting notetakers have a trust problem. To work, most of them do one of two things you probably shouldn’t love: they send a bot that shows up in the participant list and records from inside the call, or they ship your audio to someone else’s cloud for transcription. Either way, your private conversations leave your control — and good luck getting that past a security review.

Anarlog takes the opposite approach, and it’s open source under MIT.

No Bot. On Your Device. In a Format You Can Read.

Anarlog is a local-first alternative to Granola. It takes notes in your meetings without sending a bot to join the call — it listens to your device audio directly. Nothing appears in the participant list; nothing records from inside the meeting. On supported Macs, built-in transcription models run on-device, and your sessions, notes, and transcripts live in local SQLite with recordings as plain local files you can export to Markdown anytime.

The pitch, in the maintainers’ own words, is for “people who want AI meeting notes without handing their conversations to someone else’s cloud — and for anyone who needs to get a notetaker past a security review with a straight face.” That second clause is the real product insight.

The Design Choices That Matter

  • No bot joins your call. Audio is captured directly on your device.
  • Local when you choose it. On-device transcription on supported Macs; local Intelligence providers keep summaries and chat on your computer too.
  • Your data in a readable format. Local SQLite plus plain files; Markdown export whenever you want it.
  • Bring your own AI. Hosted provider, your own API key, or an OpenAI-compatible local server (Ollama, LM Studio, Unsloth). Transcription and the summary LLM are decoupled — swap one without touching the other.
  • Cloud is opt-in, not required. Hosted AI, encrypted CloudSync, and sharing exist when you want them; nothing core depends on them.
  • A real path to “yes” for IT. Source-visible enterprise components give security teams something they can actually audit.

Under the Hood

Anarlog is a Tauri v2 desktop app — React/TypeScript UI, Rust backend — with a monorepo that also ships a CLI and MCP server, an iOS/Android beta, a watchOS companion, and Tauri plugins for local STT, calendar, and export. Transcription routes through on-device models (Apple Speech / Soniox where available) or your chosen provider; summaries and chat run through whatever LLM you point it at. The active provider and model are always visible in settings — no hidden routing.

Download for macOS, Windows, or Linux at anarlog.so/download, or build it yourself from the repo.

Why It’s Worth Watching

The broader trend here is local-first AI — keeping the sensitive stuff (your audio, your transcripts) on the device while making the cloud a genuine opt-in rather than a default dependency. Meeting notes are about the most private data stream in a workday; doing them without a bot in the call and without a mandatory cloud upload is exactly the kind of design that should be the norm, not the exception.