Networkx MCP Server

What Bright-L01/networkx-mcp-server MCP server does

Bright-L01/networkx-mcp-server MCP server connects an MCP client to NetworkX-based graph analysis, with an emphasis on academic research and citation networks. It supports graph creation and editing, path queries, node and edge metadata, structural analysis, visualization, and import/export operations.

The academic functions use DOI and citation data for research workflows. The server can resolve publication metadata through CrossRef, build citation networks from seed DOIs, calculate author impact measures, inspect co-authorship patterns, identify research trends, recommend papers from citation relationships, and export citation records as BibTeX.

How it works

The server runs locally as a Python module and communicates with an MCP client over the client’s configured stdio process. Graphs are created and addressed by name, allowing subsequent tool calls to add nodes, add edges, inspect statistics, or run algorithms against the selected graph.

NetworkX supplies the underlying graph model and algorithms. The available operations cover directed and undirected graphs, shortest paths, connected components, centrality, PageRank, community detection, clustering, minimum spanning trees, cycles, coloring, matching, maximum flow, topological sorting, subgraphs, and graph composition. Citation-oriented calls use CrossRef for publication information and citation-related analysis.

Setup and configuration

Install the package with pip:

pip install networkx-mcp-server

A Claude Desktop configuration can start the server with Python’s module runner:

{
  "mcpServers": {
    "networkx": {
      "command": "python",
      "args": ["-m", "networkx_mcp"]
    }
  }
}

The README specifies Python 3.11 or newer and NetworkX 3.0 or newer. After adding the configuration, restart Claude Desktop so the tools become available. A basic verification request can create a small graph, connect nodes, and ask for a shortest path.

Tools and capabilities

Bright-L01/networkx-mcp-server MCP server groups its capabilities into several practical areas:

• Graph management: create, list, inspect, and delete graphs; add or remove nodes and edges; manage node and edge attributes.
• Graph analysis: calculate degree, betweenness, and combined centrality measures; run PageRank; find components, communities, cycles, and clustering coefficients; compute broader graph statistics.
• Optimization and structure: find shortest paths, minimum spanning trees, maximum flows, maximum-weight matchings, topological orderings, and induced subgraphs.
• Research analysis: resolve DOIs, construct citation networks, assess author impact, inspect collaboration patterns, detect trends, recommend papers, and export BibTeX.
• Data and visualization: import edge-list graphs from CSV, export graphs as JSON, and create PNG visualizations using multiple layouts.

Limitations and notes

The project is aimed at academic research and citation analysis, but it also exposes general-purpose NetworkX operations. CrossRef-dependent workflows require access to that external service; the provided setup does not show a separate API key or credential configuration. The README does not document persistence behavior, so graph storage should not be assumed to survive process restarts. The documented client setup specifically covers Claude Desktop, while compatibility with other MCP clients is not established by the provided material.