FastMCP Server For Surgical Queries Against A Vault Knowledge Graph (NetworkX No
Run precise queries against your Obsidian vault’s NetworkX node-link graph so agents answer “what connects to X?” without loading every note.
Overview
graph-query-mcp is a MCP server for the Build phase that runs targeted queries against a vault knowledge graph stored as NetworkX node-link JSON.
What is this MCP server?
- FastMCP server for surgical queries on NetworkX node-link JSON graphs
- Consumes the same graphify-out/graph.json path contract as graph-autotagger-mcp
- Avoids dumping entire vaults into context by targeting graph structure
- stdio MCPB release v0.1.0 with pinned hash
- Complements autotagger: query first, then link or summarize
- Server version 0.1.0 stdio MCPB package
- Input format: NetworkX node-link JSON graph
- Requires GRAPH_JSON_PATH environment variable
Community signal: 2 GitHub stars.
What problem does it solve?
Agents either ignore your linked note graph or ingest huge markdown dumps because there is no structured way to query vault relationships.
Who is it for?
Builders with graphify exports who want graph-native Q&A and traversal inside Claude Code or Cursor.
Skip if: Users without a generated graph.json, teams needing SQL over business data, or anyone wanting real-time Obsidian file watchers.
What do I get? / Deliverables
After pointing at graph.json, your agent can answer relationship and neighborhood questions on the graph without reading every note file.
- Graph-structured answers and traversals callable from your agent
- Lower-token vault exploration focused on links between notes
Recommended MCP Servers
Journey fit
Structured graph queries support technical docs and architecture notes during build, but the same graph supports research synthesis and post-launch runbook navigation. Docs shelf fits because the tool operates on exported vault graphs used for documentation and thinking systems, not production databases.
How it compares
Local NetworkX graph query MCP, not Elasticsearch, not Pinecone RAG, and not the wikilink suggestion server (use autotagger for that).
Common Questions / FAQ
Who is graph-query-mcp for?
Solo developers who maintain an Obsidian vault graph export and want MCP tools for precise structural queries instead of full-note retrieval.
When should I use graph-query-mcp?
Use it when planning features, writing docs, or debugging operations and you need the agent to follow note graph edges before opening specific files.
How do I add graph-query-mcp to my agent?
Place graphify-out/graph.json on disk, export GRAPH_JSON_PATH as its absolute path, install the v0.1.0 MCPB stdio bundle, and register the server in your MCP settings.