Smart Jelly

Product Introduction

1. Definition: Smart Jelly is a private, all-in-one personal AI productivity application for macOS. It is a native desktop application that functions as a unified AI agent platform, integrating features like contextual chat, system-wide dictation, voice commands, meeting assistance, task management, and automated web agents, all powered by a single, persistent, and local-first memory system.
2. Core Value Proposition: It exists to eliminate the need for multiple, siloed AI subscriptions by providing a single, cohesive AI assistant that learns from a user's entire digital work context. Its primary value is a private, unified AI memory that allows every feature—from dictation to document creation—to operate with deep contextual awareness of the user's projects, communications, and commitments, thereby increasing efficiency and reducing cognitive load.

Main Features

1. Unified Contextual Memory: This is the core technical architecture. Smart Jelly creates an encrypted, local vector database on the user's Mac that continuously indexes and semantically understands text from permitted applications (e.g., email, Slack, browsers, notes). It uses on-device processing for privacy, sending only minimal, sanitized context to cloud AI models (or a user's own API key) when a query is made. This persistent memory allows features to reference past conversations, promises, and project details.
2. System-Wide Dictation & Voice Commands: Unlike basic OS dictation, Smart Jelly's dictation is context-aware. It works in any text field (⌃⌥ shortcut) and uses the local memory to correctly spell proper nouns (names, project codenames) and adapt to the user's writing style. The Voice feature (⌥A) allows for natural language queries from within any app, with the AI able to answer verbally or type a response directly into the active text field, leveraging the unified memory for relevant context.
3. End-to-End Meeting Intelligence: This is a multi-stage feature. Before: It analyzes past communications with meeting attendees to generate predictive pre-meeting notes. During: It provides discreet, on-screen "cards" with relevant facts pulled from memory without joining the call as a bot. After: It automatically generates structured meeting notes and, critically, identifies and logs "Open Loops"—action items and promises made by the user.
4. Open Loops & Agent System: "Open Loops" is an automated task management system. It scans the unified memory (chats, emails, calls) to surface commitments ("send document to X," "follow up on Y"). It then tracks these items, reminds the user, and allows manual tick-off upon completion. The Agents feature allows the creation of semi-autonomous AI workers that perform web-based tasks (e.g., research, booking) in a dedicated browser view, showing each step and requiring user approval before final actions like payments.

Problems Solved

1. Pain Point: AI Tool Fragmentation and Cost. Professionals often subscribe to multiple single-purpose AI tools (e.g., ChatGPT for chat, Otter.ai for transcription, various SaaS for task tracking), each operating in isolation, leading to high cumulative costs and redundant data entry.
2. Target Audience: Knowledge Workers and Mac Power Users. This includes project managers, consultants, executives, developers, and writers who conduct numerous meetings, manage complex communication across apps, and need to track deliverables. It specifically targets privacy-conscious users and those frustrated with context-switching between disparate tools.
3. Use Cases: A project manager can prepare for a stakeholder call using AI-generated predictions based on all past Slack and email threads with that person; during the call, quickly ask "What was the Q3 budget we discussed?" via voice command and get an instant answer; after the call, have notes and action items auto-generated and tracked. A writer can use intelligent dictation across Google Docs, Scrivener, and email that knows their specific terminology and style.

Unique Advantages

1. Strengths & Limitations (Pros & Cons):

   • Pros:
     • Deep Workflow Integration: The single memory model is a significant technical advantage, creating a compound intelligence effect where each feature enhances the others.
     • Privacy-First Architecture: Local, encrypted memory storage and optional use of personal API keys (OpenAI, Anthropic) is a major differentiator in the AI assistant space.
     • Mac-Native Efficiency: System-wide keyboard shortcuts and deep macOS integration offer a seamless, app-agnostic user experience.
   • Cons:
     • Platform Lock-in: Exclusively for macOS (Apple Silicon, 16GB+ RAM), excluding Windows and Linux users.
     • Invite-Only / Early Stage: As of late 2026, access is gated, and some features like full local AI are marked "Coming Soon," indicating it's still maturing.
     • Learning Curve: The density of features and the concept of a "unified memory" may require an adjustment period compared to simpler, single-task tools.

2. Key Alternatives & Differentiation:

   • vs. ChatGPT Plus + Separate Tools: The primary alternative is a stack of individual subscriptions. Smart Jelly consolidates these into one app with a shared memory. Unlike ChatGPT, which has no persistent memory of your work context, Smart Jelly's answers are grounded in your actual communications. Differentiation: Unified memory, local data control, and integrated meeting/task features.
   • vs. Rewind.ai: Both offer a "record everything" local memory for Mac. Rewind focuses on universal search and recall. Smart Jelly is more prescriptive, building specific productivity features (dictation, meeting help, agents) on top of its memory. Differentiation: Smart Jelly is feature-action oriented (do things), while Rewind is primarily recall-oriented (find things).
   • vs. Microsoft Copilot for M365: Copilot is deeply integrated into the Microsoft ecosystem (Word, Teams, Outlook). Smart Jelly is app-agnostic, working across Safari, Chrome, Slack, Mail, etc., and is not tied to a specific SaaS suite. Differentiation: Cross-platform app compatibility, one-time purchase/subscription model vs. enterprise license, and stronger emphasis on user privacy with local storage.

Frequently Asked Questions (FAQ)

1. Is Smart Jelly secure and private? Where is my data stored? Yes, privacy is a core design. All your data—emails, messages, notes—is stored in an encrypted database locally on your Mac. When processing requires cloud AI, only the minimal text necessary for the query is sent, after automatically redacting sensitive information like passwords and credit card numbers.
2. How does Smart Jelly's "one memory" work technically? The app acts as a privileged, local indexer. With your permission, it reads text content from selected applications (excluding sensitive apps like banking). It processes and stores this data locally in a semantic format. When you ask a question or use a feature, it retrieves relevant context from this local index to inform the AI's response or action.
3. Can I use Smart Jelly with my own OpenAI or Anthropic API key? Yes. You can bypass Smart Jelly's cloud subscription entirely by entering your own API key in Settings. This allows you to use all app features while paying directly to your AI model provider (e.g., OpenAI for GPT-4), often providing more flexibility and potentially lower costs for high-volume users.
4. Does Smart Jelly record my meetings or join calls as a participant? No. It does not record audio or video, nor does it join calls as a bot or participant. For meeting assistance, it uses its local memory to provide context and displays helpful information on your screen only. It generates notes based on the information you have access to before and after the call.
5. What are the system requirements for Smart Jelly? It requires a Mac with Apple silicon (M-series chip) running macOS 14 (Sonoma) or later. The company recommends a minimum of 16GB of unified memory for optimal performance, especially for future fully local AI features.