zkasuran

Memfold — AI skill for Claude Code

AI community

One common memory for every AI coding tool and agent framework.

How to install Memfold

This entry records only its repository, not the path inside it, so there is no exact command to give. Open zkasuran/memfold and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Memfold does

One common memory for every AI coding tool and agent framework. Compile rules and memory out to CLAUDE.md, AGENTS.md, Cursor, Copilot, Gemini and more, or serve it live over MCP and a local daemon. Local-first, with SDKs for TypeScript, Python, Go and Rust.

Alternatives in AI

CLI Continues — resume any AI coding session in another tool — Claude Code, Copilot, Gemini, Codex, Cursor 1.5k ★
Agmsg — Cross-vendor messaging for CLI AI coding agents — let Claude Code, Codex, Gemini & Copilot talk to each other 1.5k ★
Memory Forge Rs — Stop resetting satisfying AI chats — edit the memory instead 474 ★

README

memfold

One common memory for every AI coding tool and agent framework.

[![CI](https://github.com/zkasuran/memfold/actions/workflows/ci.yml/badge.svg)](https://github.com/zkasuran/memfold/actions/workflows/ci.yml) Source-available (Apache-2.0 for the spec and SDKs, FSL-1.1 for the daemon and CLI)

You keep your project's rules and decisions in one place. memfold stores them in a local database and projects them into whatever AI coding tool you happen to be using: it compiles them into each tool's native rules file, serves them live over an MCP server and exposes them through a local REST daemon with SDKs for four languages. Write a fact once and Claude Code, Cursor, Copilot, Codex, Gemini, Cline and Aider all read the same thing.

The problem

Every AI coding tool reads from its own file. Claude Code reads `CLAUDE.md`. Codex and a growing list of tools read `AGENTS.md`. Cursor reads `.cursor/rules/*.mdc`, Copilot reads `.github/instructions`, Gemini reads `GEMINI.md`, Cline reads `.clinerules`, Aider reads `CONVENTIONS.md`. There are more than twenty of these conventions.

So you end up copying the same rules into each file by hand. They drift apart the moment one is edited. Nothing carries across when you switch tools. And because these files get committed, secrets pasted into them leak into git history. memfold removes the copying, the drift and the leaks by keeping one source of truth and generating the rest.

Try it

A real session, start to finish. No setup beyond Node 20+.

$ npx memfold init
memfold initialised
  store    ./.memfold/memory.db
  scope    project
  node id  96631630
  vectors  on
  tools    none detected

$ memfold add "Use tabs for indentation, width 4" --type preference
remembered 01M3BK22WVEJ7BHX5XWXA8QA75 (preference/project)

$ memfold add "We chose Postgres over MySQL for JSONB and partial indexes" \
    --type decision --title "Database choice"
remembered 01M3BK23WZ6KKQPEVVFCCQQTFV (decision/project)

$ memfold