Common questions about SuperLocalMemory V2, answered.

SuperLocalMemory V2 is an intelligent local memory system for AI assistants. It stores your conversations, code decisions, and project context locally, so AI assistants like Claude can remember everything about you and your projects.

Yes, 100% free. MIT license. No usage limits. No credit systems. No "free tier" restrictions. Use it commercially if you want.

No! SuperLocalMemory V2.1.0 is universal and works with 11+ IDEs:

• MCP Integration: Cursor, Windsurf, Claude Desktop, Continue.dev, ChatGPT Desktop, Perplexity, Zed, OpenCode, Antigravity
• Skills: Claude Code, Continue.dev, Cody
• CLI: Aider, any terminal
• Python API: Custom integrations

See all supported IDEs →

100% on your local machine at ~/.claude-memory/. Nothing is ever sent to any cloud service. No telemetry, no analytics, no data collection.

Yes. Your data:

• Never leaves your computer
• Is stored in a local SQLite database
• Has no network connectivity
• Requires no API keys or accounts

Yes, by default. Since no data leaves your machine, there's no third-party data processing to worry about. You have complete control over your data.

Yes. SuperLocalMemory is ideal for:

• Enterprise environments
• Classified projects
• Healthcare (HIPAA)
• Financial services
• Air-gapped systems

No. Pattern learning happens entirely locally using basic frequency analysis. No external AI services, no API calls, no internet required.

SuperLocalMemory V2.1.0 has a 7-layer universal architecture:

1. Layer 7: Universal Access — MCP + Skills + CLI (works everywhere)
2. Layer 6: MCP Integration — Model Context Protocol for 11+ IDEs
3. Layer 5: Skills Layer — 6 universal slash-commands
4. Layer 4: Pattern Learning — Learns your preferences
5. Layer 3: Knowledge Graph — Auto-discovers relationships
6. Layer 2: Hierarchical Index — Tree structure for navigation
7. Layer 1: Raw Storage — SQLite + full-text search

All layers share the same local database - no duplication, no sync issues.

Full architecture explanation →

1. Extracts key terms from your memories (TF-IDF)
2. Calculates similarity between memories
3. Groups similar memories into clusters (Leiden algorithm)
4. Auto-names clusters based on content

Example: It discovers "JWT", "OAuth", and "session tokens" are all related to "Authentication" — even if you never tagged them.

Pattern learning analyzes your memories to detect:

• Framework preferences ("React: 73% confidence")
• Coding style ("Performance over readability: 58%")
• Testing approaches ("Jest preferred: 65%")

You can feed this to Claude to get personalized suggestions.

Yes! Create isolated contexts:

superlocalmemoryv2:profile create work
superlocalmemoryv2:profile create personal
superlocalmemoryv2:profile create client-acme
superlocalmemoryv2:profile switch work

Each profile has completely separate memories, graphs, and patterns.

SQLite only (by design). Benefits:

• Zero configuration
• No server to run
• Portable (single file)
• Reliable and fast
• Works everywhere

Yes. Your data is in ~/.claude-memory/memory.db. You can:

• Copy the SQLite file
• Query it with any SQLite tool
• Write custom export scripts

ChatGPT memory:

• Is cloud-based (privacy concerns)
• Has limited capacity
• Doesn't work with Claude
• No knowledge graphs
• No pattern learning
• No multi-profile

No. Those are note-taking apps. SuperLocalMemory is specifically designed for:

• AI assistant context
• Automatic relationship discovery
• Pattern learning
• Code/development workflows

Universal integration across 11+ IDEs:

• MCP Server - Native integration with Cursor, Windsurf, Claude Desktop, Continue.dev, and 7+ more
• 6 Universal Skills - Slash-commands for Claude Code, Continue.dev, Cody
• Universal CLI - slm command works in any terminal
• Auto-Configuration - Zero manual setup for major IDEs

MCP (Model Context Protocol) is Anthropic's protocol for connecting AI assistants to external tools. SuperLocalMemory's MCP server lets AI naturally use your memory without slash commands.

Example:

You: "Remember that we use FastAPI for APIs"
Claude: [Uses remember tool automatically] ✓ Saved

Learn more →

Skills are slash-commands that work across multiple IDEs:

• /slm-remember - Save memory
• /slm-recall - Search
• /slm-status - System health
• /slm-build-graph - Rebuild graph
• /slm-list-recent - Show recent
• /slm-switch-profile - Change profile

All skills use the same local database as MCP and CLI.

Learn more →

Auto-configured (run install.sh):

• ✅ Claude Desktop
• ✅ Cursor
• ✅ Windsurf
• ✅ Continue.dev (VS Code)

Manual setup available:

• ChatGPT Desktop, Perplexity AI, Zed, OpenCode, Antigravity, Cody, Aider

Full setup guide →

Yes! All three methods use the same SQLite database at ~/.claude-memory/memory.db. No duplication, no sync issues.

Save with MCP → Search with CLI → View with Skills → All work on the same memories.

No, 100% backward compatible. All v2.0 commands still work. Your existing memories are preserved. Nothing breaks.

Unlimited. Tested up to 5,000+ memories with no issues. The graph engine has a configurable limit (default 5,000) for performance.

• Full-text search: ~45ms (3.3x faster than v1)
• Graph queries: <100ms
• Pattern lookup: <50ms

No! The knowledge graph automatically discovers relationships. Tags are optional for additional organization.

• After adding 10+ new memories — rebuild for best results
• Weekly — if you add memories regularly
• It's fast — <2 seconds for 100 memories

python ~/.claude-memory/graph_engine.py build

Yes, 100% offline. No internet connection required for any feature.

Add to PATH:

export PATH="${HOME}/.claude-memory/bin:${PATH}"

Install optional dependency:

pip install scikit-learn

1. Check if memory was added: superlocalmemoryv2:list
2. Try exact phrase search
3. Rebuild FTS index (rare): restart the app

More troubleshooting →

See Contributing for guidelines. Areas we need help:

• Performance optimizations
• Graph visualization UI
• Additional pattern categories
• Documentation improvements

Yes! See Roadmap for upcoming features.

GitHub Issues

GitHub Discussions

• ⭐ Star on GitHub
• ☕ Buy Me a Coffee
• 💸 PayPal
• 💖 GitHub Sponsors

Varun Pratap Bhardwaj - Solution Architect and Original Creator

SuperLocalMemory V2 is built as an open-source alternative to expensive cloud-based memory services like Mem0 and Zep.

• GitHub: github.com/varun369
• License: MIT (free for commercial use)
• Support: Buy me a coffee

← Back to Home

Created by Varun Pratap Bhardwaj