sandraschi/advanced-memory-mcp
If you are the rightful owner of advanced-memory-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
Advanced Memory MCP is a Model Context Protocol server that integrates personal knowledge management with Claude Desktop, offering features like knowledge graphs, zettelkasten note-taking, and experimental Claude Skills integration.
Advanced Memory MCP
A Model Context Protocol server that integrates personal knowledge management with Claude Desktop. Features include knowledge graphs, zettelkasten note-taking, and experimental Claude Skills integration.
New in v1.0.0b3: Experimental Claude Skills export and unified database architecture.
Quick Start | | |
✨ Tool Modes
PORTMANTEAU MODE (Default): 15 clean, organized tools - Perfect for Claude Desktop FULL MODE (Opt-in): 56 individual tools - For testing/development only
Result: No tool explosion! Clean, usable interface.
See for details.
Features
1. Claude Skills Format Conversion (Experimental)
# Export zettelkasten → Claude Skills format
adn_export("claude_skills", export_path="~/claude-skills/")
# Import Claude Skills → Advanced Memory notes
adn_import("claude_skills", source_path="~/anthropic-skills/")
Bidirectional format conversion between zettelkasten notes and Anthropic's Claude Skills format. Conversion tools are functional. Deployment to Claude interfaces varies (claude.ai/API verified, Claude Desktop pending verification). Part of the emerging Claude Skills ecosystem with growing community repositories.
2. Reference Library (Experimental)
87+ curated reference templates across 12 categories for systematic learning:
- Developer: Python, Git, Docker, CI/CD, Clean Code (30+ templates)
- DevOps: Kubernetes, Infrastructure as Code, Observability (15+ templates)
- Data Scientist: ML, MLOps, Python for Data Science (10+ templates)
- Researcher: Research Methods, Literature Review, Critical Thinking (12+ templates)
- Product Manager: Strategy, Analytics, Metrics (8+ templates)
- Plus: Entrepreneur, Creative, Writer, UX Designer, Knowledge Worker, AI, Philosophy
Note: These are comprehensive reference documents (1000-5000 words), not classic atomic zettelkasten notes. See for the distinction. Classic zettelkasten support (atomic notes) planned for v1.1.
3. Cursor IDE Compatible - Tool Mode Selection
Problem: Most MCPs have 40+ tools → breaks Cursor's 50-tool limit Solution: Choose between 11 portmanteau tools or full 50+ tools
Portmanteau Mode (11 tools - Cursor compatible)
adn_content → write, read, view, view_rendered, edit, move, delete
adn_project → create, switch, list, status, sync
adn_export → pandoc, docsify, html, pdf, skills!
adn_import → obsidian, notion, joplin, skills!
adn_search → search everywhere
adn_knowledge → analytics, research, bulk ops
adn_navigation → explore, recent, context
adn_editor → notepad++, typora integration
adn_zettelmaker→ generate, expand, suggest
adn_inbox → file drop processing
help → documentation
Enable portmanteau-only mode (for Cursor IDE):
{
"mcpServers": {
"advanced-memory": {
"env": {
"ADVANCED_MEMORY_PORTMANTEAU_ONLY": "true"
}
}
}
}
11 tools total = fully compatible with Cursor IDE Zero functionality loss - all features available through portmanteau tools
See for details.
4. Unified Database Architecture
- Single global database (
~/.advanced-memory/memory.db) - Project isolation via
project_id - Fast sync (2000+ notes in seconds)
- Configurable file indexing (all files or .md only)
Quick Start
Installation
Step 1: Install the package
pip install advanced-memory-mcp
Step 2: Configure your MCP client
Cursor IDE or VS Code (One-Click)
# Generate deeplink for Cursor
advanced-memory deeplink cursor
# Generate configuration for VS Code
advanced-memory deeplink vscode
# Interactive setup wizard
advanced-memory setup
Claude Desktop
Option 1: MCPB Package (Recommended for Claude Desktop)
- Download
advanced-memory-mcp.mcpbfrom Releases - Open Claude Desktop → Settings → Extensions
- Drop the
.mcpbfile - Configure project path in Extensions UI
Option 2: Manual Configuration
Add to claude_desktop_config.json:
{
"mcpServers": {
"advanced-memory": {
"command": "python",
"args": ["-m", "advanced_memory.mcp.server"]
}
}
}
Or generate the configuration:
advanced-memory deeplink claude-desktop
Option 3: Windows Bootstrap (No MCPB)
Use this when you cannot install the .mcpb package (e.g., Windsurf/Cursor on Windows).
Prerequisites: Node.js 18+, Git for Windows, Python 3.11+, uv.
# Clones to D:\Dev\repos\advanced-memory-mcp by default
npx --yes github:sandraschi/advanced-memory-mcp/scripts/bootstrap/windows
# Custom target location
npx --yes github:sandraschi/advanced-memory-mcp/scripts/bootstrap/windows -- --target C:\Work\mcp
# Also emit Cursor/Windsurf/Claude config templates
npx --yes github:sandraschi/advanced-memory-mcp/scripts/bootstrap/windows -- --generate-configs
What it does:
- Verifies dependencies (Git, Python, uv)
- Clones or updates
advanced-memory-mcp - Runs
uv sync,uv run ruff check ., and a skills validation smoke test - Optionally writes ready-to-use config templates into
bootstrap-configs/
An npm-published bootstrapper will follow; the script above is ready for npm publish when we are.
📖 Complete Installation Guide: |
First Steps
You: "Create a new project for my research notes"
Claude: Created project "research-notes"
You: "Generate starter zettelkasten for developer topics" (experimental)
Claude: Created 30+ interconnected notes on Python, Git, Docker, etc.
You: "Export my notes as Claude Skills" (experimental)
Claude: Exported 30 skills to ~/claude-skills/
Testing & Diagnostics
pwsh ./scripts/testing/run-all-tool-exercisers.ps1Runs the four portmanteau exerciser suites (core tools, import/export, skills, health/status) and stops at the first failure.- Flags:
-SkipHeavy(skip archive export/import),-SkipNetwork(skip Wikipedia/arXiv calls),-SkipPackaging(skip skills ZIP packaging). Combine flags as needed. - Individual Python scripts live in
scripts/testing/if you need to probe a single tool group.
Core Capabilities
Claude Skills Format Conversion (Experimental)
Export to Skills format:
adn_export("claude_skills", export_path="~/claude-skills/")
# Creates proper SKILL.md files for claude.ai, API, or community sharing
Import from Skills format:
adn_import("claude_skills", source_path="~/anthropic-skills/",
destination_folder="skills/anthropic")
# Imports community skills into your knowledge base
Status: Format conversion fully functional. Deploy to claude.ai (paid), Claude API, or share via GitHub. Claude Desktop integration pending verification.
Resources:
- - How to use
- - Complete ecosystem tracking
- Anthropic Skills Repo - Official examples
- Skills Spec - Format specification
- Claude Skills Docs - Official support
Intelligent File Sync
Index everything or markdown only:
{
"index_all_files": true // Code repos: true, Pure notes: false
}
Archive old content without deletion:
- Folders with
-backup-,.obsolete,-archivedautomatically skipped - Preserves files but excludes from indexing
- Perfect for keeping history without clutter
Multi-Format Export
All exports work out of the box! ✨ No paths needed!
adn_export("pdf") # → Desktop/advanced-memory-exports/pdf/
adn_export("pandoc", format_type="docx") # → Desktop/advanced-memory-exports/pandoc/
adn_export("docsify") # → Desktop/advanced-memory-exports/docsify/
adn_export("claude_skills") # → Desktop/advanced-memory-exports/claude_skills/
adn_export("html") # → Desktop/advanced-memory-exports/html/
Smart defaults: Exports go to your Desktop (or specify custom path) Zero setup: Everything auto-installs on first use PDF: Pure Python (weasyprint) - no LaTeX needed! ⚡ DOCX/EPUB: Pandoc auto-downloads (~100MB, one-time)
Documentation:
Import from Anywhere
Prerequisites: Requires existing exports/vaults from respective applications.
adn_import("obsidian", source_path="~/obsidian-vault/") # Requires: Obsidian vault
adn_import("notion", source_path="notion-export.zip") # Requires: Notion export
adn_import("joplin", source_path="~/joplin-export/") # Requires: Joplin export
adn_import("evernote", source_path="export.enex") # Requires: Evernote ENEX file
adn_import("claude_skills", source_path="~/skills/") # Requires: Skills repository
External Tools:
- Obsidian - Local-first note-taking
- Notion - Collaborative workspace
- Joplin - Open-source notes
- Evernote - Note organization
Documentation:
📖 Documentation
Getting Started
- - Step-by-step setup
- - Get productive in 5 minutes
- - Common issues
Core Features
- - Bidirectional conversion (experimental, )
- - 87+ learning templates (experimental, )
- - Cursor IDE compatibility (10 tools)
- - Unified global database
Technical & Development
- - Architecture, dependencies, build system, CI/CD
- - Extend and customize
- - System design
- - Feature verification status
Use Cases
For Development Teams
Centralize team knowledge in a structured format:
# Document team standards in Advanced Memory
# Export as Claude Skills (experimental)
adn_export("claude_skills", export_path="~/team-skills/", source_folder="team/standards")
For Researchers
Research-focused templates include:
- Research Methods Overview
- Systematic Literature Review
- Critical Thinking
- Data Analysis Fundamentals
- Academic Writing
Templates are experimental and coverage varies by domain.
For Solo Developers
Integration options for Cursor IDE:
- Index your codebase (all files)
- Generate developer zettelkasten templates (experimental)
- Export as Skills for Claude (experimental)
Provides context-aware AI assistance with your knowledge base.
🏗️ Architecture Highlights
Unified Database (v1.0.0b3)
- Single global database:
~/.advanced-memory/memory.db - Project isolation: via
project_id(not separate databases) - Fast sync: 2000+ notes indexed in seconds
- Portable: Export/import entire system as archive
Portmanteau Tools
40+ individual tools consolidated into 10 comprehensive tools:
| Portmanteau Tool | Consolidates | Operations |
|---|---|---|
adn_content | 6 tools | write, read, edit, move, delete, view |
adn_project | 8 tools | create, switch, list, status, sync, delete, set_default, get_current |
adn_export | 9 tools | pandoc, docsify, html, joplin, pdf, archive, skills |
adn_import | 6 tools | obsidian, notion, joplin, evernote, archive, skills |
adn_search | 5 tools | notes, obsidian, notion, joplin, evernote |
adn_knowledge | 9 tools | analytics, research, bulk ops, validation |
adn_navigation | 5 tools | context, recent, list, status, sync_status |
adn_editor | 5 tools | notepad++, typora, canvas, read_content |
adn_zettelmaker | 6 tools | generate, expand, suggest, connect, analyze |
adn_inbox | 4 tools | status, process, info, watch |
Result: Full functionality in 10 tools (Cursor IDE: ✅ Compatible!)
What's New in v1.0.0b3
Claude Skills Format Conversion (Experimental)
- Export: zettelkasten → Claude Skills format (SKILL.md)
- Import: Claude Skills → Advanced Memory notes
- Bidirectional metadata preservation
- Compatible with claude.ai, Claude API, emerging community ecosystem
- Claude Desktop deployment pending verification
Database Architecture Fixes
- Fixed
ADVANCED_MEMORY_HOMEdefaulting to~/.advanced-memory/ - Unified global database (removed per-project databases)
- Fixed MCPB config path issues
- Project isolation via
project_id
File Sync Improvements
index_all_filesconfig option (index all files or .md only)ARCHIVE_PATTERNS(skip backup/obsolete folders)- Enhanced diagnostic logging
MCPB Portmanteau-Only
- MCPB exposes 10 portmanteau tools (not 50+ individual tools)
- Cursor IDE compatibility
- Full functionality via consolidated tools
Python Compatibility
- Python 3.11, 3.12, 3.13 tested
- Python 3.10 compatible
- All 1113 tests passing
Learning Resources
Zettelkasten Templates (Experimental)
87+ templates across 12 categories:
Developer (30+ templates):
- Python: Fundamentals, Type Hints, Async, Testing
- Git: Version Control, Workflows
- Docker: Containers, Compose
- CI/CD: GitHub Actions, Pipelines
- Clean Code: Principles, Patterns, OOP
- System Design: Microservices, Event-Driven, Distributed Systems
Data Scientist (10+ templates):
- Machine Learning Fundamentals
- Python for Data Science
- ML Model Deployment
Researcher (12+ templates):
- Research Methods Overview
- Systematic Literature Review
- Critical Thinking
- Academic Writing
Additional categories: DevOps, Product Manager, Entrepreneur, Creative, Writer, UX Designer, Knowledge Worker, AI, Philosophy
Template quality and coverage vary. Consider them starting points for customization.
Documentation
- - Everything you need
- - Skills export/import
- - Common issues
- - Extend and customize
🛠️ Technical Details
Built With
- MCP Protocol: Model Context Protocol for LLM integration
- FastMCP: High-performance MCP server framework
- SQLAlchemy: Robust database ORM with async support
- Pydantic: Type-safe configuration and validation
- Whoosh: Full-text search indexing
- Python-Frontmatter: YAML frontmatter parsing
Requirements
- Python 3.11+ (3.10 compatible, 3.13 tested)
- Claude Desktop or any MCP-compatible client
- Windows, macOS, or Linux
Compatibility
- ✅ Claude Desktop: Native MCP integration
- ✅ Cursor IDE: 10 portmanteau tools (fully compatible)
- ✅ Claude Skills: Bidirectional export/import
- ✅ Obsidian: Import vaults, export to Obsidian format
- ✅ Notion: Import/export workspaces
- ✅ Joplin: Import/export knowledge bases
- ✅ Evernote: Import ENEX files
Example Workflows
Workflow 1: Team Knowledge to Claude Skills (Experimental)
# 1. Document team standards in Advanced Memory
adn_content("write", identifier="Python Standards",
content="# Our Python Standards\n\n...", folder="team/standards")
# 2. Export as Claude Skills (experimental)
adn_export("claude_skills", export_path="~/team-skills/", source_folder="team")
# 3. Configure Claude Desktop: Settings → Skills → Add Directory → ~/team-skills/
Workflow 2: Import and Enhance Skills (Experimental)
# 1. Import Anthropic's official skills
adn_import("claude_skills", source_path="~/anthropic-skills/mcp-builder/",
destination_folder="skills/mcp")
# 2. Enhance with your own notes and links
# 3. Export enhanced version
adn_export("claude_skills", export_path="~/enhanced-mcp-skills/",
source_folder="skills/mcp")
Workflow 3: Zettelkasten Generation (Experimental)
# 1. Generate starter templates
adn_zettelmaker("generate", category="data-scientist", topic="machine-learning")
# 2. Customize and enhance notes
# 3. Export as Skills (experimental)
adn_export("claude_skills", export_path="~/ml-skills/",
source_folder="data-scientist")
🚀 Installation Options
For Cursor IDE and VS Code
One-click installation via deeplinks provides the fastest setup:
pip install advanced-memory-mcp
advanced-memory deeplink cursor # or 'vscode'
Deeplinks automatically configure the MCP connection in your IDE.
For Claude Desktop
MCPB Package (Recommended):
- Download
.mcpbfrom Releases - Drag into Claude Desktop Settings → Extensions
- Configure via GUI
Manual Configuration:
- Install via
pip install advanced-memory-mcp - Add configuration to
claude_desktop_config.json - See for details
For Other MCP Clients
Install the Python package and configure manually per your client's documentation:
pip install advanced-memory-mcp
advanced-memory mcp # Verify installation
📊 Project Stats
- 1113 tests passing (100% pass rate)
- 87+ reference templates (12 categories, experimental - comprehensive guides, not atomic zettelkasten)
- 10 portmanteau tools (Cursor IDE compatible)
- 54% test coverage (growing)
- Python 3.11-3.13 supported (3.10 compatible)
- Zero ruff errors (strict linting)
⚠️ Testing Status
Public Repo Notice: We're committed to transparency about feature verification.
Verified ✅:
- Core content management (write, read, edit, delete)
- Project management (create, switch, list)
- Claude Skills export/import (format conversion)
- Database architecture (global db with project isolation)
Pending Real-World Verification ⏳:
- Import/export features (Obsidian, Notion, Joplin, Evernote)
- Reference template quality and coverage
- Editor integrations (Notepad++, Typora)
Auto-Installed ✅:
- Pandoc: Downloads automatically on first export (~100MB, one-time)
- Mermaid.js: Loaded from CDN for diagram rendering
- weasyprint: Pure-Python PDF generation (no LaTeX needed!)
Optional Dependencies 🔧:
- wkhtmltopdf: Alternative PDF engine (if weasyprint has issues)
- Windows:
winget install wkhtmltopdf - macOS:
brew install wkhtmltopdf - Linux:
sudo apt install wkhtmltopdf
- Windows:
- LaTeX (MiKTeX/TinyTeX): For advanced PDF typography
External Tools (optional for specific imports) 🔧:
- Obsidian/Notion/Joplin/Evernote: Only needed if importing FROM those platforms (we just read their export files)
Everything else: ✅ Works immediately after pip install!
- See for details
We're actively testing and will update status as features are verified.
🤝 Contributing
We welcome contributions! See:
- - How to contribute
- - Development setup
- - Community standards
Quick Contribution Ideas
- 🎨 Create new zettelkasten templates for your domain
- 📝 Improve documentation
- 🐛 Report bugs or suggest features
- 🌐 Translate documentation
- 🔧 Add new export/import formats
📜 License
AGPL-3.0-or-later - Free and open source
This ensures:
- ✅ You can use it freely (personal or commercial)
- ✅ You can modify and extend it
- ✅ You must share improvements (keeps it open source)
🙏 Credits
Built with inspiration from:
- Anthropic Claude Skills - Agent skills framework
- Zettelkasten Method - Knowledge management methodology
- Model Context Protocol - LLM integration standard
- FastMCP - High-performance MCP framework
🔗 Links
- Documentation:
- Releases: GitHub Releases
- Issues: Bug Reports & Features
- Claude Skills:
Quick Command Reference
# Content Management
adn_content("write", identifier="Title", content="...", folder="notes")
adn_content("read", identifier="Title")
adn_content("edit", identifier="Title", operation="append", content="...")
# Project Management
adn_project("create", project_name="research", project_path="~/Documents/research")
adn_project("switch", project_name="research")
adn_project("list")
# Claude Skills (Experimental)
adn_export("claude_skills", export_path="~/claude-skills/")
adn_import("claude_skills", source_path="~/anthropic-skills/")
# Zettelkasten Generation (Experimental)
adn_zettelmaker("generate", category="developer", topic="python-core")
# Search
adn_search("notes", query="async programming")
# Export
adn_export("pandoc", export_path="output.pdf", format_type="pdf")
adn_export("docsify", export_path="docs/")
Why Advanced Memory?
Key characteristics:
- Open source (AGPL-3.0)
- Local-first (files on your computer, works offline)
- MCP integration (designed for Claude Desktop)
- Cursor IDE compatible (via portmanteau tools)
- Extensible (import/export from multiple sources)
Experimental features:
- Claude Skills export/import
- Reference template library (comprehensive learning guides)
- Classic zettelkasten (atomic notes) planned for v1.1
Local-first knowledge management with MCP integration
• • • Report Bug