MCP Memory Visualizer

Graph visualization tools for exploring and analyzing Claude's memory data.

Graph visualization utilities for exploring and analyzing Claude's memory data captured byAnthropic's Memory MCP server.

Simply visit:https://dzivkovi.github.io/mcp-memory-visualizer/

# Install dependencies pip install -r requirements.txt # Run static analysis python visualize_memory.py # Run interactive Python version python visualize_memory_interactive.py

C:\Users\[username]\AppData\Local\npm-cache\_npx\[hash]\node_modules\@modelcontextprotocol\server-memory\dist\memory.json

⚠️ Warning:This location is temporary and gets wiped during npm cache clears or package updates.

Always configure a persistent location using theMEMORY_FILE_PATHenvironment variable in your Claude Desktop config:

{ "mcpServers": { "memory": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-memory"], "env": { "MEMORY_FILE_PATH": "C:\\Users\\[username]\\Documents\\claude-memory\\memory.json" } } } }

- C:\Users\[username]\Documents\claude-memory\memory.json
- C:\Users\[username]\AppData\Roaming\claude-memory\memory.json
- C:\claude-memory\memory.json(requires admin rights)

Note:Create the directory first and use double backslashes (\\) in Windows paths for proper JSON escaping.

The repository includes a demomemory.jsonfile with realistic but fictional data showcasing:

- 16 entitiesacross 9 different types (person, technology, project, etc.)
- 25 relationshipsforming a connected knowledge graph
- Complex connectionsbetween AI research, enterprise systems, and academic collaboration
- Varied node sizesfrom 1 to 10 observations

- Drag & Dropfile upload
- Searchentities and observations
- Interactive Graphwith physics simulation
- Detail Panelshowing observations and relationships
- Auto-layoutwith zoom controls
- Privacy-firstdesign with clear messaging

Python Static Analysis (visualize_memory.py)

- Network statistics (nodes, edges, connected components)
- Centrality analysis (most connected entities)
- Redundancy detection (similar entities, sparse nodes)
- High-resolution graph visualization (300 DPI)
- Detailed terminal analysis output

Python Interactive (visualize_memory_interactive.py)

- Browser-based interactive visualization
- Hover tooltips with full entity details
- Physics-based node positioning
- Zoom, pan, and node dragging
- HTML export for sharing

These tools work withmemory.jsonfiles in JSONL format (one JSON object per line):

{"type": "entity", "name": "Python", "entityType": "technology", "observations": ["Used for data analysis", "Popular ML language"]} {"type": "relation", "from": "Python", "to": "Data Science", "relationType": "used_in"}

- D3.jsfor powerful data visualization
- Force-directed graphlayout
- Client-side processingfor privacy
- Responsive designfor all screen sizes

- NetworkXfor graph analysis
- Matplotlibfor static visualization
- PyVisfor interactive HTML output
- Force-directed algorithmsfor natural clustering

Feel free to extend these tools with additional features:

- Export formats(GraphML, GEXF, JSON)
- Filtering options(entity types, date ranges)
- Advanced metrics(betweenness centrality, clustering coefficients)
- Memory editingcapabilities

Built for exploring Claude's memory data fromAnthropic's Memory MCP server.

Score CLAUDE.md/AGENTS.md, validate clarx-manifest.json, generate CI workflows, and pull repo scan findings into Cursor, Claude, or Grok.