M3C Tools

Multi-Modal-Memory-Capturing Tools — macOS CLI & menu bar app

View the Project on GitHub kamir/m3c-tools

M3C Tools — Multi-Modal Memory Capture

A native macOS toolkit for capturing multimodal observations (text + audio + image) and uploading them to an ER1 personal knowledge server. Built in Go with native Cocoa UI via cgo.

Quickstart

brew install portaudio
pip install openai-whisper
git clone https://github.com/kamir/m3c-tools.git && cd m3c-tools
make install
cp .env.example ~/.m3c-tools.env   # edit with your ER1 credentials
make menubar                        # launch the menu bar app

What’s in the box?

Component What it does
Menu Bar App macOS menu bar app with 4 capture channels, Observation Window, ER1 upload
CLI m3c-tools transcript — fetch YouTube transcripts, manage imports, retry queue
Transcript Library Pure Go port of youtube-transcript-api (no API key needed)
Whisper Integration Local speech-to-text via whisper CLI subprocess
ER1 Client Multipart upload to ER1 knowledge server with offline retry queue
Audio Import Batch import from a folder with SQLite tracking and bulk re-processing

Capture Channels

All channels flow through the unified Observation Window pipeline:

Capture → Preview + Record → Whisper Transcribe → Tag Editor → Store / Cancel
Channel Trigger Captures
A — YouTube Paste video URL/ID Transcript + thumbnail + voice comment
B — Screenshot Menu item Screenshot + voice note (uses clipboard if present)
C — Impulse Menu item Interactive region capture + quick voice note
D — Audio Import Menu item Batch audio files from preconfigured folder

Each observation becomes a multimodal ER1 document containing text, audio, and image with tags and metadata.

Configuration

Copy .env.example to ~/.m3c-tools.env and set at minimum:

ER1_API_URL=https://your-er1-server:8081/upload_2
ER1_API_KEY=your-api-key
ER1_CONTEXT_ID=your-context-id

See the Getting Started guide for the full configuration reference.

Documentation: