Claude Code Memory & Checkpoint Kit
Claude Code forgets everything between sessions.Every new conversation starts blank. It forgets your naming conventions, the architecture decisions you already made, your "never deploy without asking" rule, and — worst of all — what it was doing right before the context window filled up and compacted.This kit fixes both problems with a system you can set up in about five minutes.What you get • MEMORY.md — a compact index Claude reads at the start of every session, so it instantly knows what it knows. • A memory/ folder of structured notes (YAML frontmatter + [[wikilinks]]) for your project's facts, decisions, conventions, rules, and preferences. • A CLAUDE.md that tells Claude how to read, write, and maintain the memory — automatically, without you nagging it. • checkpoint.mjs — snapshots your in-flight work before context runs out, so the next session resumes exactly where you left off. • archive-memory.mjs — keeps the memory lean by retiring old notes safely (dry-run first, never deletes, never touches critical rules). • A ready-to-copy template and generic sample notes so you're not starting from a blank page.Why it's built this way • Index + compression keeps startup cheap even after hundreds of notes. • Immutable-edit discipline stops the #1 failure of memory systems: contradictory, rotting notes. • Checkpoints turn "I lost my place" into "read the last checkpoint and continue."Zero dependencies. Node 18+ and a text editor. Works on Windows, macOS, and Linux. Works with Claude Code, and the plain-Markdown notes work with any AI coding tool that reads your repo.Who it's for • Claude Code power users tired of re-explaining their project every session • Solo builders running long, multi-session projects with an AI • Anyone building toward an "AI-run solo company" workflow who needs the agent to remember decisions and hand off work reliablyBuy once, use in every project you own.
Get it → plainbuilt.gumroad.com