anarlog – Open-Source AI Meeting Notetaker That Runs Fully On-Device

TL;DR

TL;DR: anarlog is a self-hosted, MIT-licensed AI meeting notetaker that transcribes and summarizes meetings entirely on your machine — no cloud APIs, no accounts, no data leaving your device.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: anarlog on GitHub
• Direct product site: anarlog.com (redirects to GitHub)
• License: MIT (verified via GitHub license.spdx_id)
• HN launch thread: news.ycombinator.com/item?id=44725306
• Latest release: desktop_v1.0.46 (verified via GitHub Releases API, 2026-06-18)

What Is anarlog?

anarlog (formerly Hyprnote, then briefly “char”) is an open-source AI meeting notetaker that runs entirely on-device. It records meetings, transcribes them locally, and generates markdown notes — all without sending audio or conversation data to any cloud service.

The project started as a response to workplace data privacy concerns: some companies banned certain meeting notetakers, and individuals felt uncomfortable sending sensitive meeting content to third-party APIs. anarlog solves this by keeping every step of the pipeline local.

Key claims from the README (verbatim):

“An open-source AI meeting notetaker that is local-first, privacy-first, and yours to fork. Granola, rearranged.”

“Your data, your disk. Every meeting is a .md file you can inspect, search, and sync through Dropbox, iCloud, Syncthing, or git.”

“Local transcription. Transcription runs on-device, so audio never leaves your machine.”

Setup Workflow

Step 1: Download a Release

Head to github.com/fastrepl/anarlog/releases/latest and grab the installer for your platform:

# macOS ARM (Apple Silicon)
hyprnote-macos-aarch64.dmg

# macOS Intel
hyprnote-macos-x86_64.dmg

SHA-256 checksums are published alongside each release for verification.

Step 2: Install and Join a Meeting

Open the app, join your meeting as you normally would. anarlog records and transcribes in real time. Audio stays on your machine.

Step 3: Bring Your Own LLM

After the meeting, anarlog generates a summary. You configure your own LLM provider in settings:

“Bring your own AI. Use any LLM provider, including OpenAI-compatible services and local models.”

Supported providers (from README): OpenAI, Anthropic, Gemini, OpenRouter, Ollama, LM Studio, or any OpenAI-compatible API. This means you can point it at a local Ollama instance for fully offline summarization.

Step 4: Notes Are Just Markdown Files

Every meeting produces a .md file saved to disk. You own it, you can sync it anywhere, and you can edit it with any text editor.

Deeper Analysis

Privacy Model

The privacy architecture has two layers:

1. Local transcription — Audio never leaves the device during the recording session. This is the core differentiator from cloud-first alternatives like Otter.ai or Fireflies.ai.
2. Self-hosted summarization — You configure your own LLM endpoint. If you use Ollama or LM Studio, the entire pipeline is offline.

Self-Hosting

To self-host without the desktop app:

git clone https://github.com/fastrepl/anarlog.git
cd anarlog
# Follow build instructions in README

The repo contains the full source for both the desktop app and the backend pipeline.

Open-Source Forking

anarlog is MIT-licensed. The README explicitly invites commercial use:

“Open source, MIT. Fork it, sell it, or self-host it.”

This is a meaningful differentiator from hosted alternatives that restrict API access or prohibit commercial use of outputs.

Project History

The project has gone through a few name changes:

• Started as Hyprnote (YC S25)
• Briefly rebranded to char
• Split into two projects: char (team’s current productivity app) and anarlog (open-source, local-first path)

The repo maintainers note: “anarlog is not being retired. It keeps the open-source path: MIT-licensed, forkable, self-hostable, and built for local notes you control.”

Practical Evaluation Checklist

• Local transcription — audio stays on device
• No accounts or tracking — no cloud backend, no hosted account model
• Markdown output — .md files you own and control
• Bring your own LLM — OpenAI, Anthropic, Gemini, Ollama, LM Studio, or any OpenAI-compatible endpoint
• MIT licensed — fully open-source, commercial use permitted
• Cross-platform — macOS (Intel + Apple Silicon) at launch
• Self-hostable — clone, build, run your own instance

Security Notes

• Audio data never leaves the machine during transcription (local-only pipeline)
• LLM API calls only occur if you configure a cloud provider; local Ollama/LM Studio keeps everything offline
• No cloud backend means no third-party data retention to worry about
• As with any meeting recording tool, be mindful of consent laws in your jurisdiction

FAQ

Q: Does it work offline? A: Yes, if you use a local LLM provider (Ollama or LM Studio). Cloud LLM providers (OpenAI, Anthropic, etc.) require an internet connection.

Q: What transcription model does it use? A: The README does not specify the transcription engine. Check the releases and repo documentation for the current implementation details.

Q: How does it compare to Granola? A: anarlog is described as “Granola, rearranged” — the same core concept (AI meeting notes) but with a fully local, self-hosted architecture and open-source code. Granola is a hosted, proprietary service.

Q: Can I use this for work meetings on macOS? A: Yes, the desktop app supports macOS (Intel and Apple Silicon). You bring your own LLM API key if using a cloud provider.

Q: Is the code commercial-friendly? A: Yes, MIT license means you can fork, sell, or self-host without restrictions.

Conclusion

anarlog fills a specific niche: AI meeting notes with a strict local-first, privacy-first architecture. For teams that cannot or will not send meeting content to third-party APIs, this MIT-licensed open-source tool provides a credible alternative to hosted services like Granola, Otter.ai, or Fireflies.ai.

The ability to use any OpenAI-compatible LLM endpoint — including fully local options like Ollama — makes it adaptable to both cloud-heavy and air-gapped environments. Download the latest release from the GitHub releases page to get started.