Char Is the Open Source Meeting Notepad That Keeps Your Data Off Someone Else's Server

The Macro: Every Meeting Notepad Wants Your Data on Their Server

I have a problem with the way meeting transcription tools work. Otter.ai records your meetings and sends the audio to their servers. Fireflies does the same. Granola, which has gotten a lot of attention recently, takes your notes and your transcript and processes everything in the cloud. These products work well. I am not disputing that. But they all require you to trust a third-party company with the full audio and text of every conversation you have at work.

For most people, that is fine. For a lot of people, it is not. If you work in healthcare, legal, finance, defense, or any regulated industry, sending meeting audio to a third-party server creates compliance headaches that range from annoying to career-ending. HIPAA does not care that Otter has a BAA template on their website. Your compliance officer cares that patient information was transmitted to and stored on infrastructure you do not control.

Beyond compliance, there is a philosophical question. Your meeting notes are a record of your thinking, your decisions, and your work. Should that data live on someone else’s server, in someone else’s proprietary format, behind someone else’s subscription paywall? Obsidian proved that a meaningful number of people answer “no” to that question for personal notes. The same answer applies to meeting notes, but nobody has built a good product for that use case until now.

The meeting AI market has gotten crowded fast. Granola raised $43 million. Otter has been around since 2016. Fireflies, Fathom, Read.ai, Avoma, and a dozen others all compete for the same workflow. They all use roughly the same approach. Record audio, transcribe it, summarize it, maybe extract action items. The differentiation between them is thin. They compete on features, integrations, and pricing. None of them compete on data ownership.

The Micro: Three Founders, 7,000 GitHub Stars, and Zero Cloud Dependencies

Char is an open source AI notepad for meetings that runs entirely on your device. It captures system audio, generates real-time transcripts in 45-plus languages, and produces AI summaries that combine your typed notes with the transcript. Everything stays local. Audio files, transcripts, and notes are stored as plain markdown files on your machine, compatible with any tool you already use.

John Jeong and Yujong Lee are the co-founders. Deokhaeng Lee is listed as a former founder. The team is four people, part of Y Combinator’s Summer 2025 batch. The project has over 7,000 stars on GitHub, which is a strong signal for an open source productivity tool that launched less than a year ago.

The local-first approach is not just a privacy feature. It is an architectural decision that shapes the entire product. Because everything runs on-device, there is no bot joining your meeting. Char captures system audio directly, which means the people on the other end of your call never see a “Char Notetaker has joined the meeting” notification. If you have ever been in a meeting where three different AI bots joined and everyone spent the first two minutes figuring out whose was whose, you understand why this matters.

Char offers bring-your-own-key support for both speech-to-text and LLM providers. You can use cloud APIs if you want speed, or you can run local models if you want full air-gapped privacy. It works with on-premise AI models for organizations that cannot send data outside their network. The customizable templates cover sales calls, product meetings, engineering syncs, and one-on-ones.

The user list is notable. People at Databricks, Meta, Apple, Amazon, Adobe, Disney, and Bain are using it. A testimonial from a telephysician who handles six to eight patient sessions daily highlights the compliance angle. When you are discussing patient health information on every call, local processing is not a nice-to-have. It is a requirement.

The product is Mac-only for now, with Windows coming. The download is free. The business model will likely be a paid tier for team features or premium AI capabilities, though the details are not public yet.

The Verdict

Char is attacking the meeting notes market from the one angle that none of the incumbents can easily replicate. Granola, Otter, and Fireflies are cloud-first architectures. Adding a local-first mode is not a feature toggle for them. It would require rethinking their entire infrastructure. That gives Char a structural advantage with every customer who cares about data ownership, privacy, or compliance.

The risk is that most customers do not care. The majority of knowledge workers will happily send their meeting audio to whatever server is cheapest and most convenient. Char needs to win the segment that does care, and that segment needs to be large enough to build a business around.

I think it is. Healthcare alone is enormous. Legal, finance, defense, government, consulting. Any industry with confidentiality requirements is a potential market for a local-first meeting tool. And the Obsidian comparison is instructive. Obsidian proved that local-first, markdown-based, and open can build a sustainable business in personal knowledge management. Char is running the same playbook for meeting notes.

In thirty days, I want to see download numbers and daily active usage. GitHub stars are a vanity metric. What matters is whether people use Char for their actual meetings or just star the repo and go back to Granola. Sixty days, the question is whether teams are adopting it or just individuals. A single user choosing Char is nice. A compliance team mandating it across an organization is a business. Ninety days, I want to know the monetization plan. Open source meeting tools need a revenue model, and “we will figure it out later” does not work when Granola has $43 million to spend on growth. The product is excellent. The positioning is sharp. Now they need to turn philosophy into revenue.