What did we decide? Ask your recordings and check the source
A week after a meeting, someone asks: “Why did we change that plan?” You know the answer was in a call, but which one?
We’re building VoxStudio for moments like that. It transcribes recordings so you can ask about one session or search across several. Ask “What changed between these meetings?” and you can open the cited transcript passages to see what was said and when.
You can do this in VoxStudio, or connect your library to ChatGPT or Claude via MCP and bring the recording evidence into a conversation there. Cloud MCP searches your cloud sessions; Mac MCP searches your local library while the Mac app is running. An external assistant processes the transcript passages it receives, so choose the connection and context accordingly.
Here’s a short walkthrough with a lecture example and a question across three meetings: https://voxstudio.me/blog/ask-chatgpt-and-claude-about-your-recordings
If you could ask one question across your recordings, what would it be?
Replies
Since the assistant is the one calling your MCP, I'd treat every argument it sends as untrusted text, like a form field. A model-written argument is the wrong place to learn who the user is.
In a connector I run, the user comes only from the verified token, never from a parameter.
I'm curious how the Mac version handles that. With the cloud one the token gives you the scope, but a local library behind a running app has no obvious equivalent. Is it one library per app instance, or can the assistant ask to switch?