#Japan #Tokyo #AI memory #Claude Code #EpisodicRAG

A Revolutionary Step in AI Memory Management

On November 25, 2025, Meguru Co. introduced "EpisodicRAG," a groundbreaking plugin for Claude Code, aimed at empowering AI with a hierarchical memory structure capable of retaining conversations over a staggering 100 years. This innovative solution addresses the inherent limitations of AI's short-term memory, creating an opportunity for more seamless interactions in various business contexts.

The Challenge of AI Memory

The modern AI landscape is characterized by a significant issue: the transient nature of memory between sessions. This has resulted in a host of challenges:

1. Memory Gaps in Conversations
After a conversation session concludes, the AI forgets all details. Consequently, users find themselves needing to reintroduce context in each session. For long-term projects, this can lead to repetitive and inefficient information exchanges.

2. Barriers to Effective Business Utilization
In decision-making scenarios, understanding past actions is essential. AI that lacks memory of previous deliberations can only provide surface-level assistance, rendering it less effective for nuanced conversations about strategic planning and project management.

3. Limitations of Context Windows
AI's ability to process information has its constraints. Loading entire conversation histories proves daunting both technically and financially. Thus, a mechanism to prioritize and compress significant information is critical.

These challenges have impeded the adoption of AI as a long-term partner in business environments.

Unveiling EpisodicRAG

EpisodicRAG addresses these pain points through its unique functionalities:

1. An Eight-Layer Memory Structure

EpisodicRAG can compress and integrate conversation logs into a hierarchical structure, boasting the capacity to retain key information for up to 108 years.

2. Two-Tier Digest Generation

- Provisional Digest: Created immediately after a conversation to prevent fragmentation of memory.
- Regular Digest: Consolidated and refined after a set of interactions.
This dual approach effectively avoids the pitfalls of fragmented memory.

3. Memory Inheritance Across Sessions

Upon starting each session, the plugin retrieves the latest memory files from the GitHub repository, enabling users to engage in conversation while retaining previous context.

Core Advantages of EpisodicRAG

1. Seamless Business Decision-Making
AI can retain and reference the history of decisions, discussions, and learnings, allowing conversations to naturally flow from where they left off. This is particularly beneficial in fields requiring sustained contextual awareness, such as project management and corporate strategy.

2. Synergistic Effects with Existing Skills
When combined with Meguru's previously released Custom Skills, such as General Constructor and Corporate Strategist, users can experience the efficacy of an AI that possesses expert memory capabilities.

3. Versatility and Expandability
Although designed for Claude Code, access to file systems allows various coding-support AI agents to also utilize this plugin, paving the way for collaborative growth across platforms.

4. Transparency in Operation
All memory processes are stored as text files, allowing users to verify and edit contents as needed. This transparency mitigates concerns regarding the opaque nature of AI memory systems.

Understanding Claude Plugins

Claude Plugins, developed by Anthropic, enhance the capabilities of Claude Code through easily installable features. The development of EpisodicRAG has been validated through over 250 interactive sessions with Meguru's autonomous collaborative intelligence system—Weave.

Installation Steps

To install the plugin, follow these steps:
1. Launch Claude Code from your terminal: `/claude`
2. Add the marketplace: `/marketplace add https://github.com/Bizuayeu/Plugins-Weave`
3. Install the plugin: `/plugin install EpisodicRAG-Plugin@Plugins-Weave`
4. Complete the initial setup interactively: `@digest-setup`

Once set up, simply save conversation logs as "Loop" format, and execute commands to create digests and retain memory.

Societal Significance

As AI's role in business burgeons, the inability to build relationships with AI has been a common frustration for professionals. The free release of EpisodicRAG opens doors to several societal advancements:

• - On a Personal Level: Transitioning AI from a disposable tool to a growing partner, optimizing interactions based on past dialogues.
• - For Organizations: Tackling knowledge transfer within teams to ensure continuity despite personnel changes.
• - At an Industrial Level: Enhancing productivity through deepened AI engagement and collaborative models with human partners.

Technical Background

The core technology behind EpisodicRAG is currently under patent application in Japan (Application No. 2025-198943), focusing on its hierarchical memory structure and digest generation processes.

Licenses and Disclaimer

EpisodicRAG operates under the MIT license, permitting free internal use. However, prior consultation is required for any resale or integration into products due to patent rights.

Users are encouraged to refer to original conversation logs for critical decisions, as the plugin's generated memory is subject to AI interpretation. The company disclaims any liability for damages resulting from the use of this plugin.