Voice Memo

Product Introduction

1. Definition: Voice Memo is an open-source, voice-first second brain and personal knowledge management (PKM) application. Technically, it is a full-stack web application built with React, deployed as a Cloudflare Worker, and packaged into native Android and macOS applications using Capacitor and SwiftUI.
2. Core Value Proposition: It exists to solve the problem of forgotten voice notes by automatically transcribing, analyzing, and organizing spoken thoughts into actionable items, decisions, and project notes. Its core value is providing a zero-cost, self-hosted personal assistant that listens, understands, and acts on your voice memos, ensuring no idea is lost.

Main Features

1. Multi-Platform Voice Capture: The system provides native recording capabilities across platforms. The Android app uses a Java-based background service for continuous recording with the screen off. The macOS app is a SwiftUI menu-bar application with a global keyboard shortcut. The web app is a Progressive Web App (PWA) with offline functionality and an upload queue. All recordings are queued and synced to a central Cloudflare Worker backend.
2. AI-Powered Processing Pipeline: Upon upload, each audio memo undergoes a multi-stage AI workflow. It is first transcribed (supporting English, Hindi, and Hinglish) using configurable providers like Deepgram. The transcript is then cleaned and analyzed by a Large Language Model (LLM) via a quota-aware router (prioritizing Groq, with fallbacks to Mistral) to extract structured data: summary, thoughts, action points, decisions, open questions, and reminders. This data is embedded into a vector index (Cloudflare Vectorize) for semantic search.
3. Context-Aware Organization & Syncing: The system intelligently links new memos to existing data. It identifies projects from the transcript, filing related items onto a dedicated project page that maintains an auto-updating brief. Crucially, it performs change detection: if a new memo contradicts or updates an older idea (e.g., "cancel the meeting with Alex"), the relevant old action point, decision, or reminder is automatically marked as superseded, with a clear audit trail and undo functionality.
4. Integrated Claude MCP Connector: The product includes a built-in Model Context Protocol (MCP) server, enabling direct integration with Anthropic's Claude AI. After OAuth authentication, Claude can use 14 dedicated tools to read your memos and projects, add action points and notes, and set reminders, effectively becoming a conversational interface to your second brain.
5. Offline-Capable Reminder System: Reminders identified by the AI or created manually are stored and synced to native apps. On Android, they trigger full-screen alarms over the lock screen using the system's AlarmManager. On macOS, they display as floating notifications. This system works entirely offline, with background sync to the Cloudflare Worker.

Problems Solved

1. Pain Point: The "forgotten voice note" problem. Users record ideas while mobile (walking, driving) but lack the time or system to manually transcribe, categorize, and act on them, leading to lost productivity and ideas.
2. Target Audience: The primary user is a technically-inclined knowledge worker, such as a software developer, product manager, researcher, or entrepreneur, who thinks verbally and needs a frictionless system to capture and organize spontaneous thoughts without breaking flow. Secondary users are anyone seeking a private, self-hosted alternative to commercial note-taking SaaS.
3. Use Cases: Essential for capturing meeting takeaways while commuting, verbalizing and refining product decisions, maintaining a hands-free task list, logging project decisions with context, and brainstorming complex ideas that are faster to speak than type.

Unique Advantages

1. Differentiation: Unlike simple voice recorders (Apple Voice Memos, Otter.ai) which stop at transcription, Voice Memo adds a layer of AI-driven organization and action extraction. Unlike monolithic PKM tools (Obsidian, Notion), it is voice-native, automated, and built on a serverless architecture with a true $0/month running cost on Cloudflare's free tier, with all user data stored in their own Cloudflare account.
2. Key Innovation: Its change detection and reconciliation engine is a significant technical innovation. By comparing semantic embeddings of new memos against existing open items, it can automatically close loops, update decisions, and cancel obsolete reminders, mimicking how a human assistant would update notes, which is absent in all competitor products.

Frequently Asked Questions (FAQ)

1. Is Voice Memo really free to run? Yes, if you stay within Cloudflare's generous free tier limits (100,000 daily Worker requests, 10GB R2 storage, 1M Vectorize operations). The application is architected to operate within these constraints, and the AI costs are managed using free-tier-eligible providers like Groq. You only need a payment method on file with Cloudflare to activate R2.
2. How does Voice Memo handle privacy and data security? Voice Memo is self-hosted and open-source. Your audio files and processed data reside solely in your own Cloudflare account (D1 database, R2 bucket). Audio is sent to third-party AI providers for transcription but is configured to use only providers with strict no-training policies by default, and audio files are automatically purged after 30 days.
3. Can I use Voice Memo without programming knowledge? The initial deployment requires comfort with command-line tools (Wrangler, npm) and following technical documentation. It is targeted at developers and tech-savvy users who value ownership and customization. A non-technical user would need assistance to set up their own instance.
4. What AI models does Voice Memo use? The system uses a modular, provider-agnostic architecture. It primarily uses Groq's fast LLM inference for text analysis and can fall back to Mistral.ai. For transcription, it can use Deepgram or Sarvam AI. This design allows users to plug in their preferred or most cost-effective AI APIs.
5. How does the Claude MCP integration work? The Voice Memo Worker hosts a secure MCP server endpoint. You add this server URL as a custom connector in Claude's settings. After signing in with your Voice Memo credentials via OAuth, Claude can call specific, pre-defined tools (APIs) to interact with your data, allowing you to query your notes or create tasks through conversation.