Task Bert

Product Introduction

1. Definition: Task Bert is an open-source, local-first iMessage assistant and productivity tool specifically engineered for the macOS ecosystem (macOS 13.2+). Technically, it functions as a desktop application that indexes the native Apple Messages SQLite database locally, providing a semantic search layer and task-extraction interface powered by Large Language Models (LLMs).

2. Core Value Proposition: Task Bert provides the proactive intelligence users expect from a digital assistant while maintaining a strict "Zero-Knowledge" privacy architecture. By allowing users to bring their own OpenAI or Anthropic API keys, it eliminates third-party data collection, enabling private message analysis, automated calendar scheduling, and "follow-up" nudging without ever uploading message history to a proprietary server.

Main Features

1. Local Vector and Full-Text Indexing: Upon granting Full Disk Access, Task Bert parses the user’s local iMessage database and creates an embedded index on the Mac’s storage. This allows for high-speed retrieval of historical data. Unlike standard Spotlight searches, Task Bert supports natural language queries, allowing users to ask complex questions like "What was the name of the restaurant John mentioned last Tuesday?"

2. BYO-AI Integration (OpenAI & Anthropic): The application utilizes a "Bring Your Own Key" (BYOK) model for its AI features. Users input their own API keys for Claude (Anthropic) or GPT (OpenAI). This technical setup ensures that the only external communication occurs directly between the user’s machine and the AI provider. The AI is used specifically for summarizing long conversation threads and identifying actionable "Action Items" like prospective meetings or pending tasks.

3. Integrated MCP Server (Model Context Protocol): Task Bert includes a built-in MCP server, a technical standard that allows external AI agents (such as Claude Code or other MCP-compatible IDE tools) to securely query the message index. This enables developers and power users to integrate their message history into broader automated workflows or developer environments, treating their chat history as a searchable knowledge base for external agents.

4. Automated Event and To-Do Synthesis: The application identifies "loose plans" within unstructured text and offers one-click conversion into system-level Reminders or Calendar events. This feature bridges the gap between informal chat communication and structured time management without requiring manual data entry.

Problems Solved

1. Information Retrieval Friction: Users often struggle with the "buried detail" problem where critical information (addresses, dates, specific instructions) is lost in long iMessage threads. Task Bert solves this by providing a dedicated search interface that understands context better than traditional keyword matching.

2. Target Audience:

• Privacy-Conscious Professionals: Users who want AI productivity but cannot risk uploading sensitive business communications to third-party "AI wrappers."
• Freelancers and Project Managers: Individuals who manage client expectations and project timelines through text messages and need to track "invisible" tasks.
• Power Users and Developers: Those using macOS as their primary workstation who want to leverage MCP to make their communication data accessible to their local LLM tools.

3. Use Cases:

• The "Follow-Up" Nudge: Identifying messages that ended with a question or a pending request that hasn't received a response in a set timeframe.
• Context Recovery: Quickly finding the specific terms of a verbal agreement made via text months prior.
• Plan Formalization: Converting a text that says "Let's grab coffee Friday at 10" into a formal Apple Calendar invite with a single click.

Unique Advantages

1. Differentiation from Siri and Cloud Assistants: While Siri is integrated into macOS, it lacks the deep, historical semantic search and task-extraction capabilities of Task Bert. Unlike cloud-based message aggregators, Task Bert is a "no middle man" solution; the developer never sees, stores, or processes your data.

2. Key Innovation: Read-Only Local Intelligence: Task Bert is architected as a read-only utility. It cannot delete, modify, or send messages, which serves as a critical security boundary. By combining the GPL v3 open-source license with local indexing and direct-to-provider API calls, it offers a level of transparency and data sovereignty that is unique in the AI productivity space.

Frequently Asked Questions (FAQ)

1. Is Task Bert secure for sensitive business messages? Yes. Task Bert indexes your messages locally in an embedded database on your Mac. It only communicates with OpenAI or Anthropic via your personal API keys when you explicitly use AI features like summarization. No data is ever sent to Task Bert’s own servers, and the open-source code allows for full security auditing.

2. Does Task Bert require a subscription? No. Task Bert is free to download and use. However, because you "Bring Your Own Key," you are responsible for any marginal costs associated with your usage of the OpenAI or Anthropic APIs. This usually results in a much lower cost than a standard monthly AI subscription for light to moderate use.

3. Can Task Bert delete or accidentally send iMessages? No. Technically, Task Bert is a read-only application. It does not have the permissions or the internal logic to modify your message database, delete conversations, or send messages on your behalf. It is designed strictly for organization, search, and task synthesis.