QuillHub steps out: bot-free meeting recording, a new AI layer, and a shared team knowledge base
For a long time we were "that transcription bot" — and we were fine with that, right up until we realized how much gets lost between the lines of a transcript. A transcript isn't knowledge yet. Knowledge is when, a month later, you ask "so what did we actually decide about the budget with client X?" and get an answer that links straight to the exact minute of the exact meeting.
Today we're launching exactly that. Meet QuillHub — a desktop app for recording meetings, a new layer of AI analysis, workspaces for teams, and access to all your recordings through MCP. And, crucially, all of it works within your existing subscription — same minutes, same price.
Table of contents
- Why we ditched the meeting bot
- The desktop app: how it works
- Platform support, including regional services
- The new AI layer: from transcript to knowledge
- Voice biometrics: an AI that recognizes people
- Workspaces and team collaboration
- AI chat with receipts
- Access your recordings through MCP
- Same subscription, a dramatically better product
- How to join the beta
1. Why we ditched the meeting bot
The standard way to record an online meeting today is to let a bot into the call. It shows up in the participant list with a "records this meeting" badge, and everyone starts behaving a little differently. Some people clam up, and on some calls you simply can't admit a bot for compliance reasons.
There's a technical limit too: a bot only works where there's a "room" with a link. A phone call, a conversation in a meeting room, a call on a service the bot doesn't support — all of that falls through the cracks.
We took a different path. QuillHub records your computer's system audio and your microphone directly — the way Granola does, but with the Russian-speaking market squarely in mind too. No bot in the participant list. The app simply captures what you already hear — which is why it works anywhere there's sound.
2. The desktop app: how it works
QuillHub Desktop is built on Tauri (a native app, not a heavy Electron shell) and is available for macOS 13+ and Windows 10 (2004+). Under the hood there are two independent audio streams:
- System audio. On macOS we use ScreenCaptureKit; on Windows, WASAPI loopback. This is the other participants' voices.
- Your microphone is captured on a separate stream.
Why it matters: separate streams let the AI reliably tell where you are speaking versus the other side. That's the foundation for all the speaker-level analytics that follow.
Recording works even without internet
Audio is captured locally on your device, and syncing and analysis pick up as soon as a connection is available. The app is bilingual (RU/EN), with auto-updates, a global hotkey, and automatic meeting detection.
3. Platform support, including regional services
QuillHub records audio on any platform — but we deliberately built auto-detection for popular services, including the ones Western competitors ignore:
- Google Meet, Zoom, Microsoft Teams
- Telegram, WhatsApp, Webex, Skype
- Kontur Talk (KTalk), TrueConf, Yandex Telemost
- In-browser meetings, including Yandex Browser
- Ordinary phone calls and offline conversations in a meeting room
Because the foundation is system-audio capture rather than an integration with a specific API, the list is effectively open-ended: if the sound goes through your computer, QuillHub will record it.
4. The new AI layer: from transcript to knowledge
Previously, what you got out was structured text: a title, a summary, key theses, terms, a table of contents, and paragraph-level breakdowns with timecodes. That still works. But now an entire knowledge layer sits on top. After recording, a meeting runs through a pipeline that:
Extracts entities
Deals, projects, clients, candidates, people — linked to ones already known instead of creating duplicates.
Tracks state over time
A deal's status, deadline, and next step are recorded as versions, so you see how things changed meeting to meeting.
Detects conflicts
If the budget was one number in one meeting and another in the next, the AI raises an open question.
Maintains project memory
Incrementally accumulates goals, decisions, and risks per project.
Suggests actions
Proactive nudges like "create a card for this client."
Filters out noise
Classifies the recording: work meeting, dictation, or just noise/media (the latter is discarded).
This is the key distinction: not "here's the transcript text" but structured memory with receipts.
5. Voice biometrics: an AI that recognizes people
QuillHub builds voiceprints and can recognize the same person across different meetings. For each speaker, the AI isolates speech clips and matches them against named voice profiles and anonymous clusters. A new voice starts a new cluster, even without a name. The moment you say "this is me" once, your profile propagates across past and future recordings.
The practical upshot: moments — promises, decisions, concerns — get attributed to specific people rather than an abstract "Speaker 2."
6. Workspaces and team collaboration
QuillHub is no longer a single-user tool. Now there are workspaces:
- A personal workspace is created automatically and stays with you forever.
- A team workspace is a shared space with email invites and roles (owner, admin, member).
Inside you have project folders — by client, region, or product. Recordings sort themselves into the right projects based on content and participants.
Privacy protection for teams
If a personal conversation is accidentally recorded in a work team space, the AI flags the risk, and that recording stays visible only to its author until they decide what to do with it. It won't leak to the whole team automatically.
7. AI chat with receipts
A smart AI chat sits on top of your whole knowledge base. It has two modes: across the whole workspace (ask anything about all your meetings at once) and within a specific project (focus on a single client). The chat searches entities, opens their cards with current state and history, shows a timeline of how a single field changed, and lists open questions.
Most important — answers come with citations. Every response includes clickable links to the specific entity or transcript the fact came from. It's not "the AI read this somewhere," it's "here's the source, check it yourself."
8. Access your recordings through MCP
QuillHub speaks the language of modern AI agents. We have a full MCP server (Model Context Protocol) that you can connect to Claude Desktop, Cursor, Antigravity, and any MCP-aware tool.
Create transcriptions
From a URL (including YouTube/Instagram) or by uploading a file — right from your AI assistant.
Retrieve the transcript in any format
Summary, text, segments, paragraphs, chapters, subtitles.
Manage jobs
List and check status, cancel, wait for completion, and access your account and developer docs.
Authentication is via API key or OAuth (Sign in with Supabase). If a tool understands MCP, QuillHub works with it.
9. Same subscription, a dramatically better product
No extra charge
Your minutes and subscription packages for transcription apply fully to the meeting-recording app. You don't need to buy a new plan or pay extra for desktop or the AI layer. You simply get a substantially more powerful product on the same terms, for the same price.
We think that's how it should be: existing users' loyalty is rewarded with an upgrade, not a new invoice.
10. How to join the beta
QuillHub is currently in public beta. That means we're actively gathering feedback and shipping fixes fast. Install the app, try it on your own meetings, and tell us what you liked and what you didn't — it directly shapes what the product looks like at release.
Download QuillHub for Windows and macOS
The public beta is available now. Record meetings with no bot and turn them into a structured knowledge base.
Download the appFrequently asked questions
Do I have to let a bot into the call?
Which platforms does recording work on?
Is this a paid add-on for existing users?
Which operating systems are supported?
Can I work with recordings from Claude or Cursor?
Transcription was the beginning. A knowledge base built from your meetings is where we were headed all along. Give it a try — and tell us what you think.