Social Graph

Map and analyze X/Twitter social networks. Build graphs from seed accounts, detect clusters, find mutual connections, identify influencers, and get actionable follow/engage recommendations.

Quick Start

CLI

# Build a graph from a seed account (2 degrees deep, max 500 accounts)
xactions graph build --username elonmusk --depth 2 --max-nodes 500

# List saved graphs
xactions graph list

# Analyze a graph
xactions graph analyze <graph-id>

# Get follow recommendations
xactions graph recommend <graph-id>

# Export visualization
xactions graph visualize <graph-id> --format html

Node.js

import * as graph from 'xactions/graph';

// Build graph from seed user
const g = await graph.build('elonmusk', { depth: 2, maxNodes: 500 });
console.log(g.id); // graph ID

// Analyze
const analysis = graph.analyze(g, 'elonmusk');
console.log(analysis.mutualConnections);
console.log(analysis.clusters);
console.log(analysis.influenceRanking);

// Get recommendations
const recs = graph.recommend(g, 'elonmusk');
console.log(recs.followSuggestions);
console.log(recs.engageSuggestions);
console.log(recs.safeToUnfollow);

// Export
const html = graph.visualize(g, 'html');
const d3 = graph.visualize(g, 'd3');
const gexf = graph.visualize(g, 'gexf');

Dashboard

Open dashboard/graph.html for the interactive visualization interface. Features:

• Force-directed graph layout (D3.js)
• Color-coded clusters
• Node sizing by influence score
• Click nodes to view profile details
• Filter by connection type (mutual, one-way, ghost)

API Endpoints

Build a Graph

POST /api/graph/build

{
  "username": "elonmusk",
  "depth": 2,
  "maxNodes": 500
}

For large graphs (500+ nodes), the build runs in the background and emits a graph:complete Socket.IO event when done.

List Graphs

GET /api/graph

Returns an array of saved graph metadata (ID, seed user, node count, created date).

Get Graph Data

GET /api/graph/:id

Returns full graph with nodes and edges.

Get Analysis

GET /api/graph/:id/analysis

Returns:

{
  "mutualConnections": [...],
  "clusters": [...],
  "influenceRanking": [...],
  "bridgeAccounts": [...],
  "ghostFollowers": [...],
  "orbits": { "innerCircle": [...], "outerCircle": [...] }
}

Get Recommendations

GET /api/graph/:id/recommendations

Returns:

{
  "followSuggestions": [...],
  "engageSuggestions": [...],
  "competitorWatch": [...],
  "safeToUnfollow": [...]
}

Get Visualization

GET /api/graph/:id/visualization?format=d3

Formats: d3 (JSON for D3.js), gexf (XML for Gephi), html (standalone page).

Delete Graph

DELETE /api/graph/:id

Analysis Algorithms

Mutual Connections

Finds all A↔B pairs where both users follow each other. These represent your strongest connections.

Cluster Detection

Groups accounts into communities based on follow patterns. Useful for identifying niches within your network.

Bridge Accounts

Identifies nodes that connect otherwise separate clusters. Following/engaging bridge accounts expands your reach across communities.

Influence Scoring

Computes an influence score for each node based on follower count, mutual connections, and cluster centrality.

Ghost Followers

Detects inactive or low-engagement followers — accounts that follow you but never interact.

Orbit Analysis

Categorizes your network into concentric circles:

• Inner circle — mutual connections with high engagement
• Outer circle — one-way follows and periphery accounts

Graph Persistence

Graphs are saved to ~/.xactions/graphs/ as JSON files. Each graph contains:

{
  "id": "graph-elonmusk-1706000000",
  "seed": "elonmusk",
  "createdAt": "2026-01-23T12:00:00Z",
  "nodes": { "elonmusk": { "followers": 200000000, ... } },
  "edges": [
    { "source": "elonmusk", "target": "jack", "type": "follows", "weight": 1 }
  ]
}

Functions

Export Formats

D3.js

{
  "nodes": [
    { "id": "elonmusk", "group": 1, "influence": 0.95 }
  ],
  "links": [
    { "source": "elonmusk", "target": "jack", "type": "follows" }
  ]
}

GEXF (Gephi)

Standard GEXF XML format for import into Gephi for advanced network analysis.

HTML

Standalone HTML file with D3.js force-directed visualization — open in any browser.

MCP Tools

Available via the MCP server for AI agents: