minutes
Open-source meeting transcription with local storage and AI integration
AI Summary
minutes is an open-source desktop application that transcribes meetings and voice memos locally on your own computer and saves them as structured Markdown. The software uses whisper.cpp or parakeet.cpp for speech recognition, automatically extracts action items and decisions, and offers native integration with AI assistants like Claude via the Model Context Protocol (MCP). All data remains stored locally without cloud dependency.
✓ Pros
- + Fully local and open-source (MIT license) without cloud dependency or privacy risks
- + Native integration with AI assistants via MCP protocol and structured Markdown output
- + Free to use without API keys with optional LLM summarization
✗ Cons
- − Currently only available for Mac (Apple Silicon) and Windows, no Linux version mentioned
- − Requires technical understanding for setup and configuration, not as user-friendly as SaaS alternatives
Use Cases
- → Local transcription of team meetings with automatic speaker recognition and action item extraction
- → Processing iPhone voice memos on Mac with integration into personal knowledge systems
- → Quick dictation function for notes and commands via hotkey directly to clipboard
- → Searchable meeting archive for AI assistants like Claude with 29 MCP tools for querying
Who is it for?
Developers and privacy-conscious users who want to process meeting transcriptions locally and integrate them into AI workflows.
Tags
What is minutes?
minutes is a desktop application that transcribes meetings and voice memos locally and saves the results as Markdown files on your own machine. No cloud, no API keys, no external servers. Speech recognition is handled by whisper.cpp or parakeet.cpp, both running entirely offline. The source code is available under the MIT license on GitHub.
The distinctive aspect of minutes is the combination of local storage and structured output. Every transcription lands as a searchable Markdown file in your own file system. Built on top of that is a native MCP integration providing 29 tools, which allow AI assistants like Claude to search and query the meeting archive directly.
Core features
- Local transcription via whisper.cpp or parakeet.cpp, fully offline, without any cloud connection
- Automatic extraction of action items and decisions directly from the transcript
- Speaker recognition when recording team meetings
- MCP integration with 29 tools that give AI assistants like Claude access to the meeting archive
- Dictation mode via hotkey, writing spoken text directly to the clipboard
- iPhone voice memos can be processed on Mac and integrated into a personal knowledge system
Who is minutes for?
The primary audience is developers who want to incorporate meeting transcriptions into existing AI workflows, for example as context for Claude or other MCP-compatible assistants. Privacy-conscious users benefit from the fact that all data never leaves their own device.
The setup requires a basic technical understanding. Anyone who has never configured a local AI environment will need to invest time when setting up the speech models for the first time. For users looking for a ready-made SaaS solution with a web interface, minutes is the wrong starting point.
The app currently supports Mac (Apple Silicon) and Windows. Linux support is not documented.
Context & alternatives
minutes belongs to the meeting transcription tools with a local focus, a segment dominated by cloud services like Otter.ai or Fireflies.ai. These offer more user convenience and direct calendar integration, but send audio data and transcripts to external servers.
minutes takes the opposite approach. Anyone who needs transcriptions as structured input for local AI pipelines and wants to query the meeting archive programmatically gets a feature set with the 29 MCP tools that none of the SaaS alternatives mentioned offer in this form.
