Skip to content
ML2

Yogurt: private A.I. meeting notes without a bot

Added NewCopyable prompt below

Published Last verified Released

Our verdict

The open-source, local answer to paid meeting-note apps such as Granola, with one builder-specific extra: a read-only MCP server, so Claude Code or another MCP client can pull decisions and action items from your calls into specs, tickets or follow-up emails. It launched on 11 Nov 2026 with a Show HN at 224 points, and one commenter (shireboy) reported trying it on a Microsoft Teams call where it transcribed and summarized well but did not pick up the other speaker's name, which matches the docs: speaker names and calendar come from call apps on Mac only. Limits: a days-old project (v0.3.x), an Electron app that needs 8 GB of memory and downloads about 5 GB of models at setup and up to 14 GB or more afterwards, an unsigned Windows installer, and commenters point to several similar open-source options (OpenWhispr, Muesli, Biscotti). It records without anyone joining the call, so you must tell people you are recording; consent laws vary. We have not run it ourselves.

Who it's for
Builders, founders and consultants on frequent calls who want searchable notes kept off third-party servers, and anyone who wants their coding or writing agent to read meeting context
Leverage
Bot-free call transcripts, summaries and cited answers across meetings at no subscription cost, plus meeting context your agent can query directly
Effort
Low
Date added

What it does

Yogurt (FrigadeHQ/yogurt, MIT) records your microphone and the call audio playing on your computer, so no bot joins Zoom, Teams or Meet. It transcribes live with whisper.cpp (and Apple SpeechAnalyzer on supported Macs, with a 99-language mode), then writes a Summary next to your own notes using a local open-weight model run through llama.cpp. You can ask questions across meetings and get answers linked to their source, organise meetings in folders, search titles, notes and transcripts, and export notes or transcripts as Markdown. In Settings, AI apps, you can turn on a local MCP server: your assistant app can then read meetings, notes and transcripts but cannot change them; access is off by default. Data lives in a local folder per OS and can be backed up to a ZIP. It goes online only to download models from Hugging Face, fetch Apple's optional speech assets and check GitHub for updates.

How builders use it

  • Download the Mac DMG, Windows installer or Linux AppImage from GitHub Releases and let first-run setup download the models (Windows and Linux test transcription speed first).
  • Record a call, add your own notes during it, and read the Summary and per-person next steps afterwards.
  • Turn on Settings, AI apps, copy the helper command and add it as a local MCP server in your agent, then ask it to turn the latest meeting into tickets, a spec or a follow-up draft.
  • Remember that a connected cloud agent may send what it reads to its own online model; turn access off in Yogurt when you don't need it.

What to expect

Usable notes and action items from every call without a subscription or a bot in the room, and meeting decisions your agent can pull into the work that follows.

Copy this: Install + connect your agent

# Official source: github.com/FrigadeHQ/yogurt (MIT). Needs 8 GB RAM.
# Mac (Apple Silicon, macOS 14.4+):
https://github.com/FrigadeHQ/yogurt/releases/latest/download/Yogurt-mac-arm64.dmg
# Windows 10 2004+ x64 (installer not signed yet):
https://github.com/FrigadeHQ/yogurt/releases/latest/download/Yogurt-windows-x64.exe
# Linux x64 (Ubuntu 22.04+; on 24.04+ use libfuse2t64):
sudo apt install libfuse2
chmod +x Yogurt-linux-x86_64.AppImage && ./Yogurt-linux-x86_64.AppImage

# Connect your agent: Yogurt > Settings > AI apps > turn on access,
# copy the helper command, add it as a local MCP server in your agent.

# Example ask once connected:
Read my most recent Yogurt meeting. List each decision and each action item with its owner, then draft GitHub issues for the action items that touch this repo. Quote the transcript line each one comes from.

Related discoveries

See the newest picks →
  • ToolSmall Tools, Repos & Plugins

    Headroom: compress what your coding agent reads

    Apache-2.0 proxy, library and MCP server that compresses tool outputs and logs before they reach the LLM.

    Effort: MediumPrompt#Local / BYO model#MCP

  • ToolSmall Tools, Repos & Plugins

    Repomix: pack a repo into one A.I.-ready file

    One command turns a local or remote repo into a single file you can hand to any LLM, with token counts.

    Effort: LowPrompt#Plugin / extension#MCP

  • ToolSmall Tools, Repos & Plugins

    Claude Code's official plugin directory, used safely

    Anthropic's curated marketplace of Claude Code plugins: install by name from a known source instead of random third-party repos.

    Effort: LowPrompt#Plugin / extension#MCP

Get the best new A.I. picks in your inbox

The best new agent workflows and tools we’ve vetted, with a clear verdict on each. Short, occasional, useful.

More in Small Tools, Repos & Plugins →

Free. No spam. Unsubscribe anytime.