Creator · Local-first memory for Claude Code
Second Brain
The chat ends. The memory stays.
Creator, sole builder, open source · Jul 2026 – Present · Claude Code plugin · Model Context Protocol · Markdown vault (Obsidian-compatible) · Node.js / TypeScript · Next.js / Vercel
The problem
Every AI session starts from zero. You explain the same context, the model solves it, then the window closes and the reasoning is gone. Switch models or clear the chat and the memory walks away with it.
How it works
Second Brain captures every Claude Code session into Markdown on your own machine, in an Obsidian-compatible vault. It recalls the relevant notes back into new prompts, and snapshots state before the context is compacted so nothing is lost.
It installs as a Claude Code plugin (or a one-line install.sh) and runs entirely on your disk. No cloud, no lock-in.
Why it matters
Memory becomes an asset you own, not a session you rent. The longer you work, the more the vault compounds, and it moves with you across every model and machine.
