Your AI tools have amnesia. This fixes it.

What it does

Second Brain is a personal memory server that runs on Cloudflare’s free tier. You stash notes, decisions, and project context through a terminal CLI, browser extension, Obsidian plugin, or iOS Shortcuts. When you ask a connected AI client a question, it retrieves relevant memories by semantic meaning rather than exact wording, so a query about “pricing model” can surface a note that only mentions “subscription tiers.”

The interesting bit

Instead of letting each AI app silo its own memory, the project exposes a single MCP endpoint that injects your context into Claude, ChatGPT, Cursor, Codex, or any other compatible client. Your data lives in your own Cloudflare account, and an OAuth flow means you do not have to paste tokens into every app configuration.

Key highlights

• Six memory tools: remember, append, update, recall, list_recent, and forget
• Semantic search backed by Cloudflare Vectorize using 384-dimensional cosine similarity
• One /mcp endpoint connects multiple AI clients; capture inputs range from a bookmarklet to an Obsidian sync plugin
• Runs entirely on Cloudflare’s free-tier stack: Workers, D1 SQLite, KV, and Vectorize
• Data stays under your Cloudflare account, not inside a single AI platform’s walled garden

Verdict

If you split time between Claude, ChatGPT, and Cursor and are tired of re-introducing your project context, this removes the repetition. If you already live inside one AI client with decent built-in memory, it is likely unnecessary.