humla

Open-source AI meeting notes for Mac. Records mic + system audio with no bot, transcribes on-device or via OpenAI / Deepgram / Groq, identifies speakers offline

What is humla?

Open-source AI meeting notes for Mac. Records mic + system audio with no bot, transcribes on-device or via OpenAI / Deepgram / Groq, identifies speakers offline, and writes summaries that fuse your notes with the transcript. Ask your notes and get cited answers. Tauri 2 + Rust + Swift.

Overview of humla

What it is

Humla is an open-source meeting notes application for macOS. It records microphone and system audio simultaneously, transcribes the audio, identifies speakers, and produces structured summaries that combine user notes with the transcript. The application is local-first: audio and data stay on the machine unless the user explicitly sends them to a chosen provider or sync server. It is built with Tauri 2, Rust, and Swift. The name is Norwegian for 'bumblebee'.

Key features

  • Records microphone and computer audio as two separate streams, keeping voices distinct in remote calls.
  • Transcribes using local Whisper (including a Norwegian-tuned model), OpenAI (whisper-1, gpt-4o-transcribe, etc.), Deepgram (Nova-3/Nova-2), or Groq (whisper-large-v3-turbo). Per-language provider overrides are supported.
  • Identifies speakers offline using Nemotron 3 (default, up to 8 speakers) or Community-1 (pyannote clustering, for more than 8 speakers). Speaker labels can be renamed.
  • Summarizes meetings by feeding both user notes and transcript as separate inputs to a model. Presets include Meeting, 1:1, Lecture, Interview, Brainstorm, Voice memo. Supports OpenAI and any OpenAI-compatible local server (Ollama, LM Studio, etc.).
  • Provides a Chat tab that answers questions about notes using hybrid retrieval (full-text and semantic search) and cites the sources. Can be configured to use local or cloud models.
  • Ships a Model Context Protocol (MCP) server with six read-only tools (search_notes, get_note, get_transcript, list_notes, list_folders, list_clients) for use with Claude Code, Codex, or other MCP clients. It is off by default and reads the SQLite database directly without network or authentication.
  • Includes custom vocabulary, per-note language override, folders with full-text search, click-to-edit transcript, auto-update, and system-aware light/dark theme.

Use cases

Humla is designed for taking notes during meetings, both remote (Zoom, Meet, Slack huddles) and in-person. It can be used for lectures, interviews, brainstorming sessions, and voice memos. The optional Humla Cloud add-on enables sync across devices and team workspaces, with the option to self-host the sync server.

Getting started

See the repository README for installation instructions. The application is available for download at humla.team.

License and pricing

Humla is open source under the MIT license. The core app is free and does not require an account. Humla Cloud is an optional paid add-on for syncing and team workspaces; the sync server can also be self-hosted.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/michaelwilhelmsenhumla.svg)](https://yunodb.com/tools/michaelwilhelmsenhumla?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/michaelwilhelmsenhumla?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/michaelwilhelmsenhumla.svg" alt="Featured on YunoDB" height="20"></a>