How does Diverga Memory handle AI context window limits?

It uses specialized MCP tools to maintain a 500-character 'Priority Context' summary. If the session history is compressed or lost, Claude can read this summary to instantly recover the project's core state.

What happens if I skip a required checkpoint?

Required checkpoints (marked with a red icon) act as blockers. The system requires evidence of completion before allowing the research stage to advance, ensuring methodological rigor.

Where is the research data stored?

All project states, decision logs, and checkpoints are stored locally in your repository within a structured `.research/` directory using YAML and Markdown formats.

Does it support languages other than English?

Yes, the system is designed with multi-lingual support and natively recognizes both English and Korean keywords for context triggering and status updates.

Diverga Research Memory System

Name: Diverga Research Memory System
Author: HosungYou

byHosungYou

0•

Learning & Documentation

Manages long-term research context and decision history through automated checkpoints and multi-layer persistence.

The Diverga Memory System is a specialized tool for Claude Code designed to solve context loss in complex, multi-session research projects. It utilizes a sophisticated 3-layer architecture—combining natural language triggers, task-based injection, and CLI commands—to ensure project state, research questions, and methodology remain persistent. By enforcing a structured checkpoint system and maintaining an immutable, versioned decision audit trail, it allows researchers to maintain continuity, justify methodological choices, and recover essential context even when hitting AI context window limits.

Key Features

013-Layer Context System providing automatic state recovery via keywords or agent calls

02Automated Checkpoint System with REQUIRED, RECOMMENDED, and OPTIONAL enforcement levels

03Immutable Decision Audit Trail with versioning and amendment tracking for research transparency

04MCP-powered Priority Context for resilience against context window compression and recovery

05Standardized Research Workflow support covering Foundation, Design, Planning, and Execution stages

060 GitHub stars

Use Cases

01Managing complex academic or technical research projects that span multiple days or weeks

02Maintaining a verifiable audit trail of architectural or methodological decisions for team review

03Quickly resuming deep-work sessions with automated summaries of 'where we left off'

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add hosungyou/diverga memory

For use in Claude.ai and ChatGPT

Download Skill