grafeo-mcp¶
A Model Context Protocol (MCP) server that exposes GrafeoDB to AI agents. Zero infrastructure required - the server embeds the database directly.
Overview¶
grafeo-mcp lets AI agents (Claude, Copilot, etc.) query and manipulate a Grafeo graph database through the MCP standard. It runs as a standalone executable with no separate database server needed.
Installation¶
Requires Python 3.12+ and grafeo >= 0.4.4.
Quick Start¶
Claude Desktop¶
Add to claude_desktop_config.json:
With persistent storage¶
MCP Tools¶
grafeo-mcp exposes 16 tools organized in 4 categories:
Query¶
query- Execute GQL queries with auto-normalization of Cypher syntax
Graph CRUD & Traversal¶
create_node- Create a node with labels and propertiescreate_edge- Create a relationship between nodesget_node- Retrieve a node by ID with properties and connectionssearch_nodes- Find nodes by label, with optional property filtersget_neighbors- Explore a node's neighborhood (in/out/both)get_schema- Discover labels, edge types and property keysget_stats- Database statistics (counts, memory, configuration)
Vector Search¶
vector_search- k-NN similarity search using HNSW indexesmmr_search- MMR-diversified search for RAG pipelinescreate_vector_index- Create an HNSW index on a propertyhybrid_search- Combine vector similarity with graph traversal
Graph Algorithms¶
pagerank- Compute PageRank centrality scoresshortest_path- Dijkstra shortest path between nodescommunity_detection- Louvain community detectioncentrality- Betweenness centrality and connected components
MCP Resources¶
graph://schema- Rich schema with labels, properties and edge typesgraph://stats- Database statistics (counts, memory, disk, config)graph://nodes/{id}- Individual node details and connection summary
Configuration¶
| Environment Variable | Default | Description |
|---|---|---|
GRAFEO_DB_PATH | (in-memory) | Path to persistent database file |
Transport¶
Supports stdio (default), SSE and streamable HTTP transports.
Requirements¶
- Python 3.12+
- grafeo >= 0.4.4
- mcp >= 1.20
See Also¶
- grafeo-memory includes a built-in MCP server (
grafeo-memory-mcp) that wraps the high-level memory API - extract, reconcile, search, summarize. If you need AI memory management rather than raw graph access, use that instead.
License¶
Apache-2.0