How does Engram store my project data?

Engram stores your memory data locally in a .engram/ directory within your project using a ChromaDB index, ensuring your context stays private and persistent.

Can I manually tell Claude to remember something for later?

Yes, you can use the memory_remember tool to explicitly save specific information, such as user preferences or coding style requirements, for future sessions.

What happens when I start a new Claude Code session?

You can use the memory_resume tool at the start of a session to automatically inject the last session state, active todos, and recent exchanges back into Claude's context.

What is hybrid search in the context of Engram?

Engram uses a hybrid search approach that is 70% semantic (concept-based) and 30% keyword-based (BM25), allowing it to find both similar ideas and exact code snippets.

Does Engram automatically track my decisions?

While Engram can sync session transcripts, it is best practice to use the memory_decision tool to explicitly record architectural choices and the alternatives you considered.

Engram Semantic Memory

Name: Engram Semantic Memory
Author: astrosteveo

byastrosteveo

생산성 및 워크플로

Provides persistent semantic memory across Claude Code sessions to help developers recall context, architectural decisions, and project lessons.

소개

Engram enhances Claude Code by providing a persistent, searchable memory layer that bridges the gap between development sessions. Using a local ChromaDB index, it automatically stores and retrieves conversation history, tool results, and explicit user-defined context. By employing a hybrid search strategy—combining semantic conceptual matching with keyword-based retrieval—Engram ensures that Claude can access historical project decisions, coding patterns, and 'lessons learned' even after a session has ended. This skill is particularly valuable for developers managing large-scale projects where maintaining continuity and architectural consistency is paramount.

주요 기능

0 GitHub stars
Manual memory tools for explicitly saving project constraints
Hybrid 70/30 semantic and keyword (BM25) search
Automatic session resumption and context injection
Persistent local vector storage using ChromaDB
Categorized memory types for decisions, lessons, and preferences

사용 사례

Resuming complex feature development with full context after a break
Documenting and retrieving the rationale behind architectural tradeoffs
Preventing repeated mistakes by storing and searching 'lessons learned' from past bugs

소개

주요 기능

0 GitHub stars
Manual memory tools for explicitly saving project constraints
Hybrid 70/30 semantic and keyword (BM25) search
Automatic session resumption and context injection
Persistent local vector storage using ChromaDB
Categorized memory types for decisions, lessons, and preferences

사용 사례

Resuming complex feature development with full context after a break
Documenting and retrieving the rationale behind architectural tradeoffs
Preventing repeated mistakes by storing and searching 'lessons learned' from past bugs