Hemory

Product Introduction

1. Definition: Hemory is a privacy-first, always-on audio listening and memory indexing application. Technically, it is a cross-platform (iOS, Android, Apple Watch) ambient computing tool that leverages on-device and local processing to transcribe, segment, and index spoken conversations into a private, searchable knowledge base.
2. Core Value Proposition: Hemory exists to bridge the gap between lived experience and AI utility. Its primary value is transforming ephemeral real-world conversations into structured, queryable context that can be served to AI agents via the Model Context Protocol (MCP), enabling truly context-aware AI assistance grounded in factual memory.

Main Features

1. Always-On Listening with VAD: Hemory uses advanced Voice Activity Detection (VAD) to intelligently capture only moments of speech. This conserves device battery and ensures the listening quota is only consumed during actual conversation, making "always-on" listening technically and economically feasible.
2. Automatic Conversation Segmentation & Diarization: The software automatically splits continuous audio streams into discrete "moments" and applies speaker diarization (labeling, e.g., "David," "Mike") to organize content chronologically and by participant, creating a structured timeline without manual intervention.
3. MCP (Model Context Protocol) Server Integration: This is the core technical bridge. Hemory runs as a local MCP server, exposing tools like search_memory to compatible AI clients (Claude, Cursor, VS Code with Continue, etc.). This allows AI agents to query the user's personal conversation history directly within their workflow.
4. Privacy-Centric Architecture: Audio processing is stream-based, with raw audio stored exclusively on the listening device and never in the cloud. Processed transcripts and metadata are stored privately. The upcoming self-hosted option promises full infrastructure control, appealing to enterprise and high-security users.
5. Flexible Listening Modes: Offers manual control for on-demand recording and scheduled listening (e.g., 9 AM–6 PM on weekdays), giving users precise control over their privacy boundaries and recording windows.

Problems Solved

1. Pain Point: The "context gap" for AI assistants. Current LLMs lack persistent, personal context from a user's real-life interactions, making them generic. Hemory solves this by providing a continuous feed of ground-truth data.
2. Pain Point: Manual note-taking and memory decay. Professionals lose critical details from meetings, interviews, and brainstorming sessions. Hemory automates this recall, creating a perfect, searchable transcript.
3. Target Audience: Knowledge Workers & Professionals (managers, consultants, researchers), Developers & Engineers using AI-powered IDEs (Cursor, VS Code), Content Creators & Journalists conducting interviews, and Product Managers gathering user feedback.
4. Use Cases: Automating work reports from meeting transcripts, drafting Product Requirements Documents (PRDs) from customer interviews, maintaining a passive audio journal, and querying past commitments (e.g., "What did Daniel promise last Tuesday?") directly within a coding agent.

Unique Advantages

1. Differentiation vs. Standard Recorders: Unlike simple voice memo apps, Hemory provides structured, searchable memory and direct AI integration. Unlike cloud transcription services (Otter.ai, Rev), it emphasizes on-device processing and privacy-first MCP connectivity over team collaboration features.
2. Differentiation vs. Other AI Context Tools: Compared to other MCP servers or memory systems, Hemory's unique input modality is passive, real-world audio capture. It doesn't rely on manual note entry or document uploads; it builds context automatically from ambient conversation.
3. Key Innovation: The seamless integration of continuous ambient audio capture with the emerging MCP standard for AI tooling. This creates a new category of "sensory input" for AI agents, moving beyond text and files to real-world dialogue as a primary data source.

Frequently Asked Questions (FAQ)

1. How does Hemory's pricing and "listening hours" work? Hemory uses Voice Activity Detection (VAD), so your "listening time" quota only counts minutes when someone is actually speaking. Background noise and silence are ignored. This allows the app to run in always-on mode while keeping recorded minutes manageable for pricing tiers (Starter: 20 hrs/mo, Pro: 60 hrs/mo).
2. Is Hemory private? Where is my audio data stored? Hemory is designed with privacy-by-default. Raw audio is processed as a stream and stored only on your listening device (phone/watch). It is never retained in the cloud after processing. Transcripts and metadata are stored in your private memory. A future self-hosted option will allow complete data control.
3. What is MCP, and how do I connect Hemory to Claude or Cursor? MCP (Model Context Protocol) is a standard for connecting AI applications to external data sources and tools. You connect Hemory by running a simple command in your AI client (e.g., cursor mcp add hemory). Once connected, tools like search_memory become available to the AI within your chat or IDE.
4. Can Hemory record calls and online meetings? Based on its current description, Hemory is designed for ambient, in-person conversation capture via your device's microphone. It is not marketed as a call or VoIP recording solution. Its primary use case is capturing live discussions, interviews, and meetings you physically attend.
5. What platforms does Hemory support? Hemory is currently available for iOS (including Apple Watch) and Android. Native applications for macOS, Windows, Linux, and a web client are listed as "coming soon," along with the self-hosted deployment option.