How amarodeabreu/claude-graph-memory works
Claude Graph Memory is presented as local persistent graph memory for Claude Code. Its principal indexing workflow parses project documentation and code, then writes the extracted structure to NornicDB; Claude Code is documented as querying and storing memories through MCP. The tree and excerpts expose only the two population scripts, not the hook or MCP implementations, so those documented capabilities are shown without implying their wiring is present here.
Main entry points: populate-doc-graph.py Documentation indexer · populate-code-graph.go Code indexer
Components 6
Graph indexing
- Documentation indexer
scripts/populate-doc-graph.py - Code indexer
scripts/populate-code-graph.go
Graph storage
- NornicDB
Claude Code
- Claude Code
Other components
- Project docs
- Project code
Connections 5
- Project docs → Documentation indexer: reads Markdown
scripts/populate-doc-graph.py - Documentation indexer → NornicDB: writes graph
scripts/populate-doc-graph.py - Project code → Code indexer: parses Go
scripts/populate-code-graph.go - Code indexer → NornicDB: writes graph
scripts/populate-code-graph.go - Claude Code → NornicDB: queries and stores
README.md