Second Brain for AI v2

Product Introduction

1. Definition: Second Brain for AI v2 is a self-hosted, open-source personal knowledge management (PKM) system and Model Context Protocol (MCP) server. It functions as a semantic memory layer for AI assistants, enabling persistent, cross-platform context recall.
2. Core Value Proposition: It solves AI context fragmentation by providing a single, user-controlled memory that works across Claude, ChatGPT, Cursor, Codex, and any MCP-compatible client. Its primary value is eliminating repetitive explanations and maintaining project continuity across different AI tools and sessions.

Main Features

1. Semantic Memory Graph: Memories are stored as vector embeddings in a Cloudflare Vectorize index (384 dimensions, cosine metric) and linked relationally in a Cloudflare D1 SQLite database. This allows recall by meaning, not just keywords, and enables multi-hop retrieval to follow connections between related ideas automatically.
2. Multi-Platform Capture & Sync: It features a comprehensive ingestion ecosystem. This includes native MCP tools for AI clients, a CLI, a browser extension, iOS Shortcuts, an Obsidian plugin, and a new Notion integration. The Notion sync uses an internal connection to automatically import and update shared pages, ensuring memory stays current.
3. Self-Hosted & Open-Source Architecture: The entire stack runs on Cloudflare Workers, leveraging D1, Vectorize, KV, and Workers AI. Deployment is via a one-click Cloudflare button, placing all data and infrastructure within the user's own Cloudflare account, ensuring data privacy, ownership, and no vendor lock-in.
4. Graceful Degradation & OAuth 2.0: The system includes operational resilience features. If the Vectorize index is unavailable, recall falls back to keyword search. A /health endpoint provides status. For authentication, the /mcp endpoint supports OAuth 2.0 discovery, allowing clients like Claude Desktop to securely authenticate via a browser flow without exposing the static AUTH_TOKEN in configuration.

Problems Solved

1. Pain Point: Context Amnesia Across AI Tools. Users lose project history, decisions, and preferences when switching between Claude, ChatGPT, and Cursor, forcing them to re-explain context repeatedly.
2. Target Audience: The primary users are developers, product managers, researchers, and knowledge workers who utilize multiple AI assistants for different tasks (e.g., coding with Cursor, writing with Claude, brainstorming with ChatGPT) and need a unified memory.
3. Use Cases: Essential for software developers tracking architectural decisions across coding sessions; product managers maintaining consistent feature specs and user feedback across AI tools; writers and researchers building a connected knowledge base from disparate AI interactions.

Unique Advantages

1. Differentiation: Unlike built-in memory features in individual AI apps (e.g., ChatGPT's memory), Second Brain is platform-agnostic, open-source, and self-hosted. Unlike generic note-taking apps, it is optimized for low-friction capture from AI chats and semantic retrieval tailored for AI assistant context windows.
2. Key Innovation: Its implementation of a hybrid memory graph—combining vector search for semantic similarity with a relational graph database for explicit connections—enables "associative recall." This allows the AI to traverse linked memories (hops), surfacing relevant context that a simple semantic search might miss.

Frequently Asked Questions (FAQ)

1. How does Second Brain for AI v2 handle privacy and data ownership? All data is stored exclusively in your own Cloudflare account using D1 (SQLite) and Vectorize. The software is open-source, and no data is sent to any third-party servers, ensuring complete user control and ownership.
2. What is the cost of running Second Brain for AI v2 on Cloudflare? For personal use, the application operates entirely within Cloudflare's generous free tier limits for Workers, D1, Vectorize, and KV, making it effectively free to run at a personal scale.
3. Can I use Second Brain with AI tools other than Claude and ChatGPT? Yes, any application that supports the Model Context Protocol (MCP), such as Cursor, Codex CLI, or custom clients, can connect to the /mcp endpoint and use the remember, recall, and other memory tools.
4. How does the Notion sync integration work technically? You create an internal "connection" in Notion's developer portal, share pages with it, and provide the connection's secret to Second Brain. It then syncs page content into memory, with updates processed nightly or manually triggered via "Sync now" in the dashboard.
5. What happens if my Vectorize index is deleted or fails? Second Brain v2 implements graceful degradation. The /health endpoint will report the issue, the web dashboard will show a warning banner, and the recall function will automatically fall back to a keyword-based search within your D1 database to ensure functionality remains.