What information does the memory skill track?

The skill proactively records architectural decisions, team conventions, bug fix root causes, configuration changes, and non-obvious discoveries about the codebase.

How does Engram handle session resets or compactions?

Engram uses a mandatory session summary protocol that saves the current goal, accomplishments, and discoveries to a SQLite database before compactions, allowing Claude to reload state immediately upon restart.

Does this work with other AI agents?

While optimized as a Claude Code Skill via MCP, Engram is an agent-agnostic Go binary that can provide memory capabilities to any AI coding assistant through its HTTP API or CLI.

Can I search through my past coding decisions?

Yes, Claude can perform full-text searches using the mem_search tool to retrieve specific historical observations based on keywords or topic keys.

Engram Persistent Memory

Name: Engram Persistent Memory
Author: Gentleman-Programming

byGentleman-Programming

•

1,751

•

生産性とワークフロー

Maintains long-term project context and technical decisions across AI coding sessions using a specialized SQLite-backed protocol.

Engram Persistent Memory provides a robust, agent-agnostic memory layer that allows Claude to proactively store and retrieve critical project information, including architectural decisions, bug fixes, and development conventions. By integrating a mandatory protocol for saving discoveries and session summaries, it ensures that context survives session compactions and restarts, preventing redundant questions and maintaining a continuous workflow for complex software projects. It leverages a Go-based binary with FTS5 search capabilities to provide fast, reliable access to historical session data via the Model Context Protocol (MCP).

主な機能

01Automated session summarization to preserve project state across compactions

02Persistent SQLite storage that survives across multiple AI coding sessions

03Full-text search (FTS5) for lightning-fast historical context retrieval

041,751 GitHub stars

05Structured observation categories for bugs, architecture, patterns, and preferences

06Proactive decision and convention tracking that triggers without user prompts

ユースケース

01Resuming complex feature development across multiple days without losing architectural context

02Documenting non-obvious bug fixes and root causes to prevent regression in future tasks

03Maintaining a consistent set of project-specific conventions and team coding standards

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add gentleman-programming/engram memory

For use in Claude.ai and ChatGPT

Download Skill