Second Brain for AI

Product Introduction

1. Definition: Second Brain for AI is an open-source, self-hosted memory layer and knowledge management system designed for artificial intelligence applications. It functions as a persistent, semantic memory backend for large language model (LLM) clients and MCP (Model Context Protocol) interfaces, enabling tools like Claude, ChatGPT, and Cursor to retain and recall user-specific context across sessions.
2. Core Value Proposition: This product solves the fundamental "stateless" problem of AI interactions by providing a centralized, user-controlled AI memory management system. Its core value is enabling true contextual continuity—allowing users to store decisions, project details, and preferences once and recall them via semantic search in any connected AI tool, eliminating repetitive re-explanation and creating a persistent personal knowledge base.

Main Features

1. Semantic Recall and Memory API: The system provides a REST API with core endpoints (/capture, /recall, /append, /update, /forget) for programmatic memory management. Instead of keyword matching, it uses vector embeddings and semantic search (powered by Cloudflare Vectorize) to find memories based on conceptual meaning. The /recall endpoint returns results by contextual relevance, not exact text match. The API is compatible with the MCP (Model Context Protocol), allowing seamless integration with compliant AI clients.
2. Duplicate Detection and Data Integrity: Integrated duplicate detection logic prevents redundant storage of the same information. When capturing new data, the system checks for existing semantic matches to suggest consolidation via the append or update actions, ensuring a clean and efficient personal knowledge graph. The underlying storage is built on Cloudflare D1 (a SQLite-based serverless database), ensuring transactional integrity.
3. Multi-Platform Capture and Web UI: Context can be ingested from numerous sources: a command-line interface (second-brain-cf-cli), an Obsidian plugin for note syncing, iOS Shortcuts for mobile capture, a Chrome browser extension, and a simple Web UI dashboard. This ensures the memory layer is filled from the natural points of context creation across devices and workflows.
4. Self-Hosted, Serverless Cloudflare Stack: The entire backend is deployed as a Cloudflare Worker, leveraging a free-tier-compatible stack: Cloudflare Workers for compute, D1 for relational data, Vectorize for vector storage, and Workers AI for generating embeddings. This architecture provides global low-latency access, scalability, and ensures user data remains within their own Cloudflare account, fully under their control.

Problems Solved

1. Pain Point: Context Amnesia in AI Tools. The primary problem is the context loss experienced when switching between different AI platforms (e.g., from Claude to ChatGPT) or starting new conversations. Users must repeatedly provide the same project details, coding preferences, or decision history, severely hampering efficiency and continuity in long-term projects.
2. Target Audience: This product is essential for software developers using multiple AI coding assistants (like Cursor and Claude), researchers and writers who juggle multiple projects across AI platforms, product managers documenting decisions in various tools, and any power user of LLMs who wants to build a persistent, private, and portable knowledge base.
3. Use Cases: Key scenarios include a developer recalling a previously decided architectural pattern during a code review in Cursor; a researcher finding all stored notes related to "market analysis strategies" without remembering exact file names; a team sharing a unified project context that any member can query through their preferred AI client; and capturing a fleeting idea from a mobile device via an iOS Shortcut to be referenced later in a desktop AI session.

Unique Advantages

1. Differentiation from Built-in Memory: Unlike the proprietary, platform-locked memory features within ChatGPT or Claude, Second Brain for AI is user-owned, cross-platform, and open-source (MIT licensed). It is not controlled by a single AI provider, meaning memories persist independently of any service's terms of service, data policies, or feature changes. It provides a unified memory layer, not fragmented, siloed memories.
2. Key Innovation: MCP-Native Semantic Backend. The key technical innovation is its design as a Model Context Protocol (MCP) server from the ground up. While it offers a traditional REST API, its native MCP support makes it a first-class citizen in the emerging ecosystem of interconnected AI tools. This allows any MCP-compliant client to interact with the memory layer as a standardized tool, promoting interoperability. The combination of serverless Cloudflare infrastructure with built-in semantic search provides a powerful, scalable, and cost-effective (free-tier friendly) foundation for a personal AI memory.

Frequently Asked Questions (FAQ)

1. How does Second Brain for AI's semantic search work? It uses vector embeddings generated by a Cloudflare Workers AI model. When you store (capture) or search (recall) a memory, the text is converted into a high-dimensional vector representing its meaning. Recall finds memories whose vectors are closest in meaning to your query vector, enabling conceptual matching beyond keywords.
2. Is my data private and where is it stored? Yes, data privacy is core to the design. When you self-host on Cloudflare, all data is stored in your own Cloudflare account (in D1 and Vectorize). It is not sent to the creator of Second Brain or any third party. The system is MIT licensed, allowing full audit and modification of the code.
3. Can I use this with tools other than Claude and ChatGPT? Absolutely. It is designed for any AI tool that supports the Model Context Protocol (MCP), such as Cursor. Furthermore, its REST API (/capture, /recall, etc.) can be integrated into any custom script, application, or workflow that can make HTTP requests.
4. What does "self-hosted" mean for deployment and cost? Deployment uses Cloudflare's one-click "Deploy to Workers" button. "Self-hosted" means it runs in your Cloudflare infrastructure, not on a third-party managed service. For personal use, it fits within Cloudflare's free tier for Workers, D1, Vectorize, and AI, making it potentially cost-free while you retain full control.