Chat with Your Meeting Notes via MCP

The problem: meetings disappear

You took the meeting. You might have a transcript. But when you need to recall what was said — what was decided about pricing, who owned the API migration, what the client's timeline constraint was — you're scrolling through pages of text or searching your memory.

Meeting transcripts are valuable data. But they're usually trapped in a format that's hard to query: a wall of timestamped text in an app you have to open and search manually.

MCP: making meetings queryable

Model Context Protocol (MCP) is an open standard that lets AI assistants access external data sources. Instead of copying transcript text into a prompt, the AI calls tools that search and retrieve the relevant content itself.

Resonant's MCP server exposes your meeting data as tools an AI can use:

• Search meetings. Find meetings by date, keyword, participant, or topic.
• Get transcript. Retrieve the full or partial transcript of a specific meeting.
• Get summary. Pull AI-generated summaries, action items, and key decisions.
• Query across meetings. Ask questions that span multiple meetings — “What has the client said about budget across all our calls?”

What you can ask

Once connected, you can ask your AI assistant things like:

• “What were my action items from this week's standups?”
• “What did the engineering team decide about the database migration?”
• “Summarize all meetings with Acme Corp in the last month.”
• “What concerns did the VP raise about the Q2 plan?”
• “Find every meeting where we discussed the webhook redesign.”
• “Draft a follow-up email based on today's client call.”

The AI has access to the transcript content and can reason over it — finding patterns, extracting commitments, and synthesizing information across meetings.

How it works technically

1. Resonant transcribes meetings locally on your Mac (on-device, no cloud).
2. Transcripts are indexed and stored locally.
3. Resonant runs an MCP server that exposes search and retrieval tools.
4. You configure your AI client (Claude Desktop, Cursor, etc.) to connect to the MCP server.
5. When you ask about meetings, the AI calls the MCP tools to find relevant transcripts.
6. The AI reasons over the retrieved content and gives you an answer.

Privacy model

The MCP server runs locally on your machine. Meeting audio is always local — it never leaves your Mac. When you query via a cloud AI (Claude, ChatGPT), the relevant transcript excerpts are sent to that AI for processing — similar to pasting text into a prompt. You control which meetings are queryable and which AI you use.

For fully local queries, connect a local LLM (via Ollama or similar) as the MCP client. This keeps the entire pipeline — audio capture, transcription, storage, and AI queries — on your Mac.

Compatible AI tools

• Claude Desktop — native MCP support via connectors.
• ChatGPT — MCP connectors in the App Store.
• Cursor — MCP tool integration for coding workflows.
• Windsurf — MCP support for development contexts.
• VS Code + Continue — open-source AI extension with MCP.
• Any MCP client — the protocol is open and standardized.

Related resources

• Meeting transcription webhooks— push transcripts to external systems automatically.
• Automatic meeting notetaker— how meeting capture works on Mac.
• Private meeting transcription— the privacy architecture.
• Voice for AI engineers— how developers use Resonant in their workflow.

Frequently asked questions