The Open-Source Meeting Notes Tool That Keeps Your Conversations Off the Cloud

Meetily is a free, MIT-licensed GitHub project with 12,400 stars that transcribes and summarizes meetings entirely on your own machine, replacing cloud-based tools like Otter.ai and Fireflies.ai that cost $100 to $360 per seat per year. No bot joins your call, no audio leaves your device, and the whole thing runs on Whisper or Parakeet with an Ollama-powered summary layer you control.

Meetily, an MIT-licensed open-source project sitting at 12,400 GitHub stars, runs AI meeting transcription and summarization entirely on your own computer, replacing the per-seat subscription model that tools like Otter.ai and Fireflies.ai charge $100 to $360 per year per person to operate. You open a meeting, Meetily captures both your microphone and system audio simultaneously, transcribes in real time using Whisper or NVIDIA's Parakeet model locally, and generates a structured summary using a local LLM via Ollama or a cloud API of your choosing. Nothing touches a third-party server unless you explicitly route it there.

That last part is the actual product differentiation. Otter.ai and Fireflies.ai are selling convenience, but they are also holding your meeting recordings on infrastructure you do not control. For a team in healthcare, legal, finance, or consulting, that is not a neutral tradeoff. Meetily bets that a meaningful slice of that market would rather manage its own meeting data than pay $30 a month per seat not to.

What it actually does during a meeting

Meetily captures audio using a Rust-based backend that mixes microphone input and system playback simultaneously, capturing both sides of a call at the OS level without routing through virtual devices or kernel extensions. The transcription engine runs locally using Whisper.cpp or NVIDIA's Parakeet model, which the project claims is four times faster than standard Whisper on compatible hardware. On an M-series Mac or a CUDA-capable GPU machine, the live transcript appears within a few seconds of speech.

After the meeting, Meetily passes the transcript to a summarization layer. By default it uses Ollama running locally, so the summary never leaves your machine either. You can swap in Claude, Groq, or any OpenAI-compatible endpoint for higher-quality summaries, but even then what gets sent is the text transcript, not the audio recording.

The output is a structured summary with key decisions and action items. There is no speaker diarization in the community edition yet. If you run tight meetings with clear attribution, that gap is minor. If you run sprawling discussions where knowing who committed to what matters, it is a real limitation.

The installation gap between community and Otter

Otter.ai takes forty seconds to set up. You create an account, connect your calendar, and a bot joins your next meeting. Meetily requires slightly more intention, though less than most open-source tools in this category.

On macOS, you download a DMG, drag it to Applications, and launch it. The app walks you through granting the audio permissions it needs. If you want local LLM summarization rather than a cloud provider, you install Ollama separately and pull a model. That is the full setup path for a non-technical user on a Mac. On Windows, there is an x64 setup executable in the releases. On Linux, you build from source, which is a legitimate technical lift.

The difference between the Mac experience and what you pay for at Otter or Fireflies is mostly meeting detection and a bot presence. Meetily does not auto-detect that a meeting started and join it for you. You open the app before the call and press record. For internal teams running scheduled calls, this is a trivial habit change. For salespeople who hop on impromptu calls from mobile throughout the day, it is a genuine friction point that the cloud tools handle more gracefully.

Where the privacy math actually changes the decision

The privacy case for self-hosted meeting tools is not primarily about data breaches. The more direct issue is regulatory. GDPR and HIPAA create obligations around where data is processed and stored. A law firm recording client calls through Otter is transmitting privileged conversation data to a U.S. cloud provider. A healthcare organization summarizing patient discussions through Fireflies is creating a third-party data relationship that triggers BAA requirements. These are not edge cases. They are the default experience for thousands of organizations using cloud meeting tools without thinking carefully about it.

Meetily is the alternative. The community edition is free, and the PRO tier, which adds GDPR compliance tooling and enhanced transcription, starts at $10 per user per month, roughly the same price as Fireflies Pro. For a five-person team on Otter Business ($150/month), the break-even on developer setup time is reached before month two.

Honest gaps

The community edition has no speaker diarization, which means the transcript identifies what was said but not who said it. For teams that care about action item attribution, this is a meaningful limitation. The PRO tier lists it as coming soon, but the timeline is not public.

Meeting auto-detection does not exist. You press record manually. If your team forgets to open the app before a call starts, you lose the first few minutes. Otter and Fireflies eliminate this friction entirely by joining as a bot. The tradeoff is that Meetily never discloses its presence to external meeting participants who did not consent to recording, which is either a benefit or a concern depending on your jurisdiction and ethical stance.

The Ollama summarization quality is model-dependent. Running llama3.1:8b for summaries on a machine without a GPU produces functional but not exceptional output. The summaries cover action items and key points but miss nuance in complex technical discussions. Switching to a larger local model or a cloud endpoint meaningfully improves the output, but the default experience at zero cloud cost is average.

What 12,400 stars means

The star count is not just developer enthusiasm. It reflects the specific appetite of the knowledge worker who sits in five meetings a day, pays $16.99 per month for Otter, and has thought carefully about where those recordings go. The privacy argument is not paranoid. It is a rational response to the fact that enterprise SaaS companies routinely use customer data to train their models, the legal exposure of cloud meeting recordings is real and growing, and the technical barrier to avoiding that exposure has dropped to the level of a twenty-minute setup.

Meeting tools that run locally were not viable three years ago because the models were too slow or too inaccurate to be useful. Parakeet and Whisper running on Apple Silicon changed that. What Meetily does is package that capability into a desktop application that a non-engineer can install in an afternoon.

The meetings you have been uploading to the cloud every day for the past three years are likely still there, on someone else's infrastructure, associated with your account. Meetily does not solve that retrospectively. But starting today, it gives you the option of owning the next three years of conversations instead.