JamesShi96/project-butler

What it solves

AI coding assistants often suffer from "forgetfulness" across different sessions, requiring users to re-explain project architecture, naming conventions, and previous decisions. project-butler provides a persistent memory layer that turns a project folder into a shared source of truth, allowing different AI tools (like Claude Code, Cursor, and Codex) to pick up exactly where the last session left off.

How it works

The tool implements a 7-component memory stack using plain Markdown files and JSON configs stored directly in the project root. This stack organizes information by stability:

• Stable Rules: Human-reviewed principles stored in CLAUDE.md.
• Current State: A project wiki (PROJECT.md), file organization rules (STRUCTURE.md), and a milestone changelog (UPDATE_LOG.md).
• Raw Facts: Session logs and a TODO.md checklist.
• Handoff: A session-handoff.md file that explicitly tells the next assistant how to resume.

It integrates as a native skill for Claude Code and generates compatibility files (like .cursor/rules/project-system.mdc or AGENTS.md) for other assistants, ensuring they follow the same memory protocol.

Who it’s for

Developers who use AI coding assistants and want to avoid repetitive context-setting, especially those switching between different AI tools or working on long-term projects with complex architectures.

Highlights

• Cross-Tool Compatibility: Works across Claude Code, Cursor, and Codex by using a file-based memory system.
• Coded Handoffs: Uses specific triggers like end session to save progress and continue to resume context.
• Profile-Awareness: Can create a Project Profile to manage document tiers and protected sections for stronger architectural alignment.
• Automated Maintenance: Includes features to organize files, sync wikis, and auto-check for its own updates.