claude-mem  by thedotmack

AI coding assistant memory system

Created 2 months ago
325 stars

Top 83.6% on SourcePulse

GitHubView on GitHub
Project Summary

A Claude Code plugin, claude-mem, addresses the challenge of maintaining context and project knowledge across coding sessions. It automatically captures, compresses using AI, and re-injects relevant information, enabling Claude to retain continuity even after sessions end or reconnect. This offers significant benefits for users working on complex, long-term projects by reducing the need for repetitive context-setting and improving AI recall.

How It Works

Claude-Mem operates by intercepting session events via lifecycle hooks. It captures user prompts and tool execution observations, processing them with the Claude Agent SDK to extract key learnings and generate semantic summaries. These summaries are stored in a local SQLite database and automatically injected at the start of new sessions. This layered approach, termed "progressive disclosure," allows for efficient context retrieval, mirroring human memory patterns and providing access to detailed historical data when needed.

Quick Start & Requirements

To install, run the following commands within a Claude Code session:

/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem

Restart Claude Code for context to appear automatically.

Prerequisites:

  • Node.js: 18.0.0 or higher
  • Claude Code: Latest version with plugin support
  • PM2: Process manager (bundled)
  • SQLite 3: For persistent storage (bundled)

Documentation: Browse markdown docs on GitHub

Highlighted Details

  • Persistent Memory: Context survives across sessions, including restarts and reconnections.
  • Progressive Disclosure: Layered memory retrieval (index, details, source) with token cost visibility.
  • 7 MCP Search Tools: Enables querying project history via specialized search functions like search_observations, find_by_file, and get_recent_context.
  • Automatic Operation: No manual intervention required; context is injected transparently.
  • Structured Observations: AI-powered extraction and categorization of learnings (decisions, bugfixes, features) with concept and file tagging.

Maintenance & Community

The project is authored by Alex Newman (@thedotmack). Support and issue tracking are managed via GitHub Issues. The CHANGELOG.md indicates active development.

Licensing & Compatibility

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). This strong copyleft license requires that any modifications or derivative works, especially when deployed on a network server, must also be made available under the AGPL-3.0. This license may impose significant restrictions on integration with closed-source or proprietary systems.

Limitations & Caveats

The AGPL-3.0 license presents a notable adoption blocker for commercial or closed-source environments due to its strict source-sharing requirements. The system relies on the Claude Code environment and its plugin architecture. While actively developed, recent critical fixes suggest potential for ongoing bugs or instability. Resource footprint details are not explicitly provided.

Health Check
Last Commit

23 hours ago

Responsiveness

Inactive

Pull Requests (30d)
15
Issues (30d)
22
Star History
323 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.