Earlyn

Product Introduction

1. Definition: Earlyn is a native macOS productivity application that functions as a private, on-device screen memory and meeting intelligence system. Technically, it is a local-first AI agent platform that captures, encrypts, indexes, and semantically analyzes screen content and audio data directly on an Apple Silicon Mac.
2. Core Value Proposition: Earlyn exists to solve the problem of digital amnesia and fragmented context by providing a fully private, searchable memory of everything a user sees and hears on their computer. Its primary value is enabling total recall of on-screen information and meeting details without compromising data privacy, as all processing occurs locally.

Main Features

1. Rewind (Screen Memory Timeline): This feature continuously captures and indexes the textual content of the user's screen. It operates by taking periodic screenshots (approximately once per second during activity), using on-device OCR and text extraction to convert pixels to searchable text. The system only stores a new "moment" when the on-screen text changes, optimizing storage. The interface presents a scrubbable visual timeline, allowing users to drag through a chronological history of their workday and visually locate specific words or events.
2. Ask (Semantic Search & Q&A): This is an on-device semantic search and question-answering engine over the user's captured history. It utilizes local embedding models to convert queries and stored text into vector representations, enabling search by meaning, not just keyword matching. When a user asks a question (e.g., "What was the error code from this morning?"), the system retrieves relevant moments and uses a local or user-configured cloud LLM (like Claude or OpenAI) to generate an answer, complete with citations to the source screen captures or transcripts.
3. Meetings (On-Device Transcription & Intelligence): This feature automatically detects audio from video conferencing apps (like Zoom) and offers to record and transcribe. It uses the Whisper large-v3-turbo model running locally on the Mac's Neural Engine to perform speech-to-text in approximately 100 languages, with automatic language detection. It includes speaker diarization, live translation using Apple's on-device translation frameworks, and generates post-meeting summaries with extracted action items, all processed locally.
4. Memory for AI Agents (MCP Server Integration): Earlyn acts as a local Model Context Protocol (MCP) server. With one click, it integrates with AI-powered coding environments like Claude Code, Cursor, and Windsurf, allowing those agents to query the user's personal screen and meeting history via a localhost (127.0.0.1) connection. This gives AI development tools persistent, contextual memory of the user's work.

Problems Solved

1. Pain Point: The inability to find previously viewed digital information, such as a lost error message, a shared link, or a specific comment from a past meeting, leads to significant time wasted in manual searching.
2. Pain Point: Reliance on cloud-based meeting transcription and memory services creates privacy and data sovereignty risks, as sensitive audio and screen data is transmitted to third-party servers.
3. Target Audience: Knowledge Workers & Remote Professionals: Individuals who spend their days in meetings and across numerous applications (Slack, browsers, documents) who need to recall specific details. Software Developers & Engineers: Professionals using AI-assisted coding tools (Cursor, Windsurf) who need to provide rich, historical context to their AI pair programmer. Executives & Consultants: Those who handle sensitive information in meetings and documents and require a secure, auditable memory system.
4. Use Cases: A developer debugging an issue can ask, "Show me the API error from the server logs I had open two hours ago." A project manager can get a transcript and action items from a client call in Turkish, translated live to English. A researcher can semantically search across weeks of PDFs and browser tabs for a specific concept without remembering the exact terminology.

Unique Advantages

1. Differentiation vs. Competitors: Unlike cloud-dependent alternatives, Earlyn's core architecture is local-first and private-by-design. It contrasts with Rewind.ai (which discontinued its Mac screen capture) and Microsoft Recall (exclusive to Windows Copilot+ PCs) by being a fully functional, privacy-centric solution for the macOS ecosystem. It requires no user account for its core functionality.
2. Key Innovation: The integration of multiple on-device AI models (OCR, Whisper STT, embedding models, Apple Translation) into a cohesive, efficient system that runs entirely on a Mac's Apple Silicon chipset. The "zero-server-trust" model is its foundational innovation—privacy is an architectural guarantee, not an optional setting. The local MCP server implementation is a unique technical approach to bridging personal memory with external AI agent ecosystems securely.

Frequently Asked Questions (FAQ)

1. Is Earlyn a local or cloud application? Earlyn is primarily a local application. All core functions—screen capture, text extraction, search indexing, meeting transcription, and translation—run directly on your Mac. Data is only sent externally if you explicitly choose to use a cloud AI model (like OpenAI) for Q&A or connect an external app like Slack.
2. How does Earlyn's privacy compare to Rewind or Microsoft Recall? Earlyn is architected for maximum privacy. Unlike services that process data in the cloud, Earlyn performs all processing on-device. Data is encrypted at rest using SQLCipher (database) and AES-GCM (images), with keys stored in the macOS Keychain. It is designed to never capture password fields, private browser windows, or content containing detected secrets.
3. What are the system requirements for Earlyn? Earlyn requires a Mac with Apple Silicon (M-series chip) running macOS 14 (Sonoma) or later. It does not support Intel-based Macs or Windows PCs due to its deep integration with Apple's Neural Engine and on-device ML frameworks.
4. Can I use Earlyn to search through my documents and files? Yes. Beyond screen content, Earlyn can index and perform semantic search on documents (PDFs, Word, PowerPoint, Markdown) stored in folders you explicitly designate, integrating this data into the same unified search interface as your screen and meeting memories.
5. What happens to my data after the 7-day free trial ends? When the trial concludes, the active recording of new screen and meeting data pauses. All historical data you captured during the trial remains encrypted and searchable on your Mac until it is automatically purged by the system's retention policy (e.g., after 7 days for the free tier). You can export your data at any time.