Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
henyeu247-max avatar

Claude Mem

  • Updated May 2, 2026
  • henyeu247-max/claude-mem-xai

claude-mem is a Claude Code skill in the AI & Agent Building category. Persistent memory system for Claude Code - context compression across sessions (xAI Grok only)

Key points

  • claude-mem
  • AI & Agent Building
  • AI-coding skill

Claude Mem by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add henyeu247-max/claude-mem-xai
/plugin install claude-mem@henyeu247-max

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedMay 2, 2026
Repositoryhenyeu247-max/claude-mem-xai

What it does

Persistent memory system for Claude Code - context compression across sessions (xAI Grok only)

README.md

claude-mem-xai

Multi-platform fork of claude-mem by Alex Newman

License Version Upstream

🎯 What is this?

A specialized fork of claude-mem with multi-platform and multi-provider support:

  • Multi-platform: Claude Code, Windsurf, Cursor, Gemini CLI
  • Multi-provider: xAI Grok, NVIDIA NIM (removed Claude SDK)
  • Chroma disabled by default for lower CPU usage
  • SQLite-only search mode as default
  • Windsurf Cascade Hooks integration with automatic memory context injection

🚀 Key Differences from Original

Feature Original claude-mem claude-mem-xai
AI Providers Claude, Gemini, OpenRouter, xAI xAI Grok, NVIDIA NIM
IDE Platforms Claude Code only Claude Code, Windsurf, Cursor, Gemini CLI
Vector DB (Chroma) Enabled by default Disabled by default
Search Mode Hybrid (Chroma + SQLite) SQLite-only
CPU Usage Higher (chroma-mcp subprocess) Lower
Default Model claude-sonnet-4-5 grok-4-1-fast-non-reasoning

📦 Installation

# Install via Claude Code
claude install plugin henyeu247-max/claude-mem

# Or manually
git clone https://github.com/henyeu247-max/claude-mem-xai.git
cd claude-mem-xai
npm install
npm run build

🌊 Windsurf Integration

claude-mem works with Windsurf IDE via Cascade Hooks — automatically capturing your coding sessions and injecting memory context.

Install Windsurf Hooks

# Install for current project
claude-mem windsurf install

# Install globally (all projects)
claude-mem windsurf install user

# Check installation status
claude-mem windsurf status

# Uninstall
claude-mem windsurf uninstall

How It Works

Windsurf Cascade Hooks fire at key points during your coding session:

Windsurf Event claude-mem Action
pre_user_prompt Session init + context injection
post_read_code Record file read observation
post_write_code Track file edits
post_run_command Record command execution
post_mcp_tool_use Record MCP tool usage
post_cascade_response Generate session summary
post_cascade_response_with_transcript Process full transcript

Context is automatically written to .windsurf/rules/claude-mem-context.md — Windsurf loads this as a Rule in every Cascade conversation, giving the AI memory of your past sessions.

Setup Steps

  1. Start the worker: claude-mem start
  2. Install hooks: claude-mem windsurf install
  3. Restart Windsurf to load the hooks
  4. Verify: Check Windsurf Settings → Hooks

🖥️ Cursor Integration

# Install Cursor hooks
claude-mem cursor install

# Check status
claude-mem cursor status

⚙️ Configuration

Settings are stored in ~/.claude-mem/settings.json:

xAI Provider

{
  "CLAUDE_MEM_PROVIDER": "xai",
  "CLAUDE_MEM_XAI_API_KEY": "xai-your-api-key-here",
  "CLAUDE_MEM_XAI_MODEL": "grok-4-1-fast-non-reasoning"
}

NVIDIA NIM Provider

{
  "CLAUDE_MEM_PROVIDER": "nvidia",
  "CLAUDE_MEM_NVIDIA_API_KEY": "nvapi-your-key-here",
  "CLAUDE_MEM_NVIDIA_MODEL": "openai/gpt-oss-120b"
}

Available Models

xAI Grok:

  • grok-4-1-fast-non-reasoning (default) - Fastest, no reasoning
  • grok-4-1-fast-reasoning - Fast with reasoning
  • grok-3-fast - Previous generation

NVIDIA NIM:

  • openai/gpt-oss-120b (default) - Best performance/cost ratio
  • nvidia/llama-3.3-nemotron-super-49b-v1 - NVIDIA's model
  • deepseek-ai/deepseek-r1 - DeepSeek reasoning

Chroma Vector Database

Chroma is disabled by default. To enable:

{
  "CLAUDE_MEM_CHROMA_ENABLED": "true"
}

� Performance

Mode CPU Usage Processes
Original (Chroma enabled) ~300-500 MB worker + chroma-mcp + python
This fork (Chroma disabled) ~150-200 MB worker only

�🔧 Why This Fork?

Original claude-mem is excellent, but:

  1. Single platform - Only Claude Code, no Windsurf/Cursor support
  2. Claude SDK dependency - Requires Claude subscription for AI processing
  3. Chroma enabled by default - High CPU usage from chroma-mcp subprocess

This fork adds:

  1. Multi-platform - Windsurf, Cursor, Claude Code, Gemini CLI
  2. Alternative AI providers - xAI Grok, NVIDIA NIM (no Claude subscription needed)
  3. Lower resource usage - No Chroma subprocess by default
  4. Windsurf context injection - Auto-writes memory to .windsurf/rules/

🙏 Credits

This fork is based on claude-mem by Alex Newman (@thedotmack).

All core functionality and architecture credit goes to the original project and its 78+ contributors.

📝 License

AGPL-3.0 (same as original claude-mem)

🔗 Links

🤝 Contributing

This is a personal fork for xAI/NVIDIA multi-platform use. For general claude-mem contributions, please contribute to the original project.

For improvements to this fork, feel free to open issues or PRs.


Made with ❤️ by henyeu247-max | Based on claude-mem by Alex Newman

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.