trnscrb is a macOS menu bar app designed for professionals who need private, efficient local meeting transcription. It belongs to the category of productivity tools that automate meeting notes by capturing audio and transcribing it entirely on your device using Whisper AI, with no cloud dependency. The core value proposition is that it eliminates manual note-taking while ensuring all data stays local, and it integrates seamlessly with Claude Desktop for instant, AI-powered search across all transcripts. This makes it ideal for privacy-conscious users who attend frequent video meetings and want a seamless, automated transcription workflow without worrying about data leaving their computer.
The primary problem trnscrb solves is the distraction and inefficiency of taking notes during meetings, which often leads to missed action items and fragmented information. Many users rely on cloud-based transcription services that require internet and raise concerns about data privacy and ownership. With trnscrb, there is no need to manually start recording, no cloud processing, and no account sign-up for basic transcription. It automatically detects when you are in a meeting by monitoring microphone activity and the presence of common meeting applications, then records and transcribes locally. This solves the pain point of losing important details while keeping your audio and transcripts completely under your control, which is especially critical for teams handling sensitive information.
Auto-detection and recording are the first major feature group. trnscrb listens for microphone activity lasting five seconds while detecting running processes such as Google Meet, Zoom, Slack Huddle, Microsoft Teams, or FaceTime. This dual trigger ensures that recordings only happen when you are actually in a meeting, preventing false starts. You can choose to capture only your microphone audio or, by installing the free BlackHole virtual audio driver, also capture system audio from other participants. trnscrb install automates this setup, so you do not need to manually configure virtual sound devices. The benefit is that you get complete meeting recordings without any manual effort, and the auto-detection means you never miss the start of a meeting while fumbling with a recording button.
The second major feature group is on-device transcription using the Whisper small model, optimized for Apple Silicon via Metal acceleration. This runs entirely offline, so no internet connection is required for transcription, and audio files are processed locally and deleted after transcription to conserve storage. The output is saved as plain .txt files into the ~/meeting-notes/ directory, making them universally readable and portable. trnscrb also offers speaker labels if you set up a free HuggingFace account, which the installer guides you through. The advantage of this approach is unmatched privacy—your conversation data never leaves your machine—and reliability, as transcription can happen even in remote areas with no internet. The use of the small model balances speed and accuracy, delivering results quickly on modern Mac hardware.
admin
A third feature group is the integration with Claude Desktop via the Model Context Protocol (MCP). After running trnscrb install, Claude Desktop gains a set of tools that allow it to control recordings and search transcripts. These MCP tools include start_recording, stop_recording, recording_status, get_last_transcript, list_transcripts, get_transcript, get_calendar_context, and enrich_transcript. The enrich tool sends the transcript text to the Claude API (the only time any data leaves your machine, and only if you explicitly run it) to generate a summary and action items. This integration means you can ask Claude natural language questions like "What were the action items from this morning's standup?" and get directly relevant answers from your transcripts, effectively turning your past meetings into a searchable knowledge base.
The overall workflow is simple and automated. After installation via Homebrew or pip, trnscrb install configures BlackHole, downloads the Whisper model, sets up the HuggingFace token (if needed), configures Claude Desktop MCP, and ensures the app launches at login. The app lives silently in the menu bar. When a meeting is detected, it begins recording. When the meeting ends (detected by microphone inactivity or app closure), transcription runs automatically in the background. The resulting plain text file is saved immediately, and within moments it becomes searchable via Claude Desktop queries. The user never needs to think about starting or stopping recordings unless they want to manually control the process using the MCP tools. This frictionless approach ensures every meeting is captured without disrupting the user's focus.
Concrete use cases are demonstrated effectively in the product's examples. A user can ask "What were the action items from this morning's standup?" and receive a list of tasks with owners and deadlines. Another common scenario is starting a recording on-demand by telling Claude "Start recording — I'm about to jump on a call," which initiates capture via the MCP tool. After a call, the user can say "Stop and summarise the call" to stop recording, get a transcript, and have Claude generate a summary with action items. A powerful cross-meeting search is illustrated by the query "What has Sara mentioned across all my recent meetings?" which scans all transcripts to find mentions and context. The outcome is that users save hours of manual note-taking, never lose important discussion points, and can quickly retrieve information from months of meetings.
The target audience is primarily macOS users who attend many video meetings and value privacy and productivity. This includes project managers who need to track action items, software developers who want to document technical discussions, remote team leads who coordinate across time zones, and privacy-conscious professionals who avoid cloud transcription services. The app works on Apple Silicon Macs (recommended for Metal acceleration) and Intel Macs (slower but functional). It integrates with the most common meeting platforms: Google Meet, Zoom, Slack Huddle, Microsoft Teams, and FaceTime. The tech stack relies on Whisper AI, BlackHole for system audio, and MCP for Claude Desktop integration. The product is distributed as a free command-line tool via Homebrew and pip, with no pricing tiers mentioned, making it accessible to anyone with a Mac. In summary, trnscrb delivers the most private and automated local meeting transcription experience for macOS, turning every conversation into a searchable asset without compromising data ownership.
Project managers, software developers, remote team leads, and privacy-conscious professionals who use macOS and attend frequent video meetings across platforms like Google Meet, Zoom, Slack Huddle, Microsoft Teams, and FaceTime. The app is designed for individuals who want automated, searchable meeting notes without relying on cloud services and who value data sovereignty. It is particularly useful for users of Claude Desktop who wish to integrate meeting transcripts into their AI-powered workflows for quick retrieval of information. The tool benefits anyone who needs to keep a permanent, local record of discussions and action items, especially in environments where confidentiality is paramount.