AI Meeting Assistants

anarlog9.6k★

anarlog takes meeting notes without sending a bot into the call. It records your mic and computer audio locally, transcribes it, writes the summary, and keeps everything in a local SQLite database. It's the open source answer to Granola and Otter, formerly Hyprnote, and the desktop app is MIT licensed and free. Fully local is still an Apple Silicon story. The on-device transcription models need an M-series Mac. On Windows, Linux or an Intel Mac you bring a speech-to-text key, point it at a Deepgram-compatible server, or pay for the cloud. Windows and Linux are in beta. Pro is $14 per person per month or $140 per person per year for cloud transcription, a cloud model, encrypted sync and shared team workspaces. Enterprise adds SSO and SCIM at custom pricing. Solo on an M-series Mac: free. Small teams sharing notes: Pro. Larger orgs: Enterprise. The catch: with no bot, nobody else on the call is told they're being recorded, and in all-party-consent states like California getting permission is on you; its meeting-chat disclosure is off by default. Usage analytics stay on until you switch them off, and web search in chat goes through anarlog's servers even with local models.

quill3.9k★

Quill records a call and transcribes it without anything leaving your Mac. It's a tiny menu-bar app: one click records your mic and the call audio as separate tracks, and on stop a local speech model writes a timestamped transcript. Not the Quill.js editor; same name, different project. MIT licensed and free. There's no app to download. You clone the repo, build it with Apple's Swift toolchain, then run sudo ./scripts/install-macos. It needs macOS 15 or later, and the first transcription pulls about 600 MB of model files. Transcription is English-only and fast on Apple Silicon. Speaker labels are just "me" and "them", because your mic is one track and everything the Mac plays is the other, notification dings included. No summaries; an on_stop hook hands the folder to your own script. Solo: free, and cheaper than Otter for raw transcripts. Teams get no sharing or sync. The catch: it records everyone on the call with no bot and no notice to them, and in all-party-consent states like California getting permission is on you. It's also a young, mostly one-developer project with no tagged releases, and nothing has merged since September 27 while outside fixes sit in open pull requests.

minutes1.6k★

Minutes turns your meetings, calls and voice memos into Markdown files your AI assistant can search. It records on your machine, transcribes locally, can tag speakers, and writes each conversation to ~/meetings/ with decisions and action items in the frontmatter. Then Claude Code, Codex, Cursor or any MCP client can tell you what you decided last month. MIT licensed Rust, and the author promises no paid tier. There's a signed desktop app for Apple Silicon Macs and Windows, plus Homebrew, cargo install and a prebuilt Linux CLI. The Mac app transcribes with Parakeet; other installs use Whisper. Summaries are off by default: your agent summarizes over MCP, or you point it at Ollama, a Claude or Codex subscription, or an API key. Solo: free, best if you live in Claude Code or Obsidian. Small teams: sharing means sharing a folder. Larger teams on Granola Business at $14 per seat get shared workspaces and admin controls Minutes doesn't attempt. The catch: with no bot, nobody else on the call is told automatically, and in all-party-consent states like California getting permission is on you, though Minutes reminds you by default. It's also pre-1.0 and ships releases almost daily, so expect churn.

cue1.5k★

cue is a floating AI window for macOS, Windows and Linux that watches your screen, listens to your meetings and suggests what to say next, hidden from screen shares. It's a GPL-3.0 clone of Cluely. The code is free; the default download runs on publik API, a metered AI gateway, unless you paste your own key. Download a release, grant mic and screen access, and you're in. Your own key works with OpenAI, Anthropic, Gemini, Groq or a local Ollama server, and local whisper.cpp transcription keeps audio on your machine. publik bills each request at half the model's list price, with plans from $8/month. publik's own site links its GitHub to the cue author's account, and publik pays app makers a cut of its markup. Solo: free with your own key. Teams: nothing here for you. The catch: consent. cue transcribes everyone on the call, with no bot and no notice, and keeps their words in a plain-text file. On the default path their words and your screenshots go through publik's servers, and setup sends your computer's name. Where the law requires everyone's consent to record, doing that without telling people can be illegal. Tell them.