Menu
Advanced Hive Mind collective intelligence system for queen-led multi-agent coordination with consensus mechanisms and persistent memory
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Evolve this harness with Darwin Mode — frozen model, evolving harness (real, sandboxed, safety-gated).
Turn a feature request into a minimal, file-level implementation plan before any code.
Create and run custom background analysis workers with composable phases. Use when you need automated code analysis, security scanning, pattern learning, or API documentation generation.
Quantum-resistant, self-learning version control for AI agents with ReasoningBank intelligence and multi-agent coordination
Train and deploy neural networks in distributed E2B sandboxes with Flow Nexus
Comprehensive Flow Nexus platform management - authentication, sandboxes, app deployment, payments, and challenges
| name | hive-mind-advanced |
| description | Advanced Hive Mind collective intelligence system for queen-led multi-agent coordination with consensus mechanisms and persistent memory |
| version | 1.0.0 |
| category | coordination |
| tags | ["hive-mind","swarm","queen-worker","consensus","collective-intelligence","multi-agent","coordination"] |
| author | Claude Flow Team |
| hooks | {"pre":"echo \"🧠 Hive Mind Advanced activated\"\nif [ -d \"/workspaces/ruvector/.claude/intelligence\" ]; then\n cd /workspaces/ruvector/.claude/intelligence\n INTELLIGENCE_MODE=treatment node cli.js pre-edit \"$FILE\" 2>/dev/null || true\nfi\n","post":"echo \"✅ Hive Mind Advanced complete\"\nif [ -d \"/workspaces/ruvector/.claude/intelligence\" ]; then\n cd /workspaces/ruvector/.claude/intelligence\n INTELLIGENCE_MODE=treatment node cli.js post-edit \"$FILE\" \"true\" 2>/dev/null || true\nfi\n"} |
Master the advanced Hive Mind collective intelligence system for sophisticated multi-agent coordination using queen-led architecture, Byzantine consensus, and collective memory.
Integrates with RuVector's Q-learning and vector memory for improved performance.
CLI: node .claude/intelligence/cli.js stats
The Hive Mind system represents the pinnacle of multi-agent coordination in Claude Flow, implementing a queen-led hierarchical architecture where a strategic queen coordinator directs specialized worker agents through collective decision-making and shared memory.
Queen-Led Coordination
Worker Specialization
Collective Memory System
Majority Consensus Simple voting where the option with most votes wins.
Weighted Consensus Queen vote counts as 3x weight, providing strategic guidance.
Byzantine Fault Tolerance Requires 2/3 majority for decision approval, ensuring robust consensus even with faulty agents.
# Basic initialization
npx claude-flow hive-mind init
# Force reinitialize
npx claude-flow hive-mind init --force
# Custom configuration
npx claude-flow hive-mind init --config hive-config.json
# Basic spawn with objective
npx claude-flow hive-mind spawn "Build microservices architecture"
# Strategic queen type
npx claude-flow hive-mind spawn "Research AI patterns" --queen-type strategic
# Tactical queen with max workers
npx claude-flow hive-mind spawn "Implement API" --queen-type tactical --max-workers 12
# Adaptive queen with consensus
npx claude-flow hive-mind spawn "Optimize system" --queen-type adaptive --consensus byzantine
# Generate Claude Code commands
npx claude-flow hive-mind spawn "Build full-stack app" --claude
# Check hive mind status
npx claude-flow hive-mind status
# Get detailed metrics
npx claude-flow hive-mind metrics
# Monitor collective memory
npx claude-flow hive-mind memory
Create and Manage Sessions
# List active sessions
npx claude-flow hive-mind sessions
# Pause a session
npx claude-flow hive-mind pause <session-id>
# Resume a paused session
npx claude-flow hive-mind resume <session-id>
# Stop a running session
npx claude-flow hive-mind stop <session-id>
Session Features
The Hive Mind builds consensus through structured voting:
// Programmatic consensus building
const decision = await hiveMind.buildConsensus(
'Architecture pattern selection',
['microservices', 'monolith', 'serverless']
);
// Result includes:
// - decision: Winning option
// - confidence: Vote percentage
// - votes: Individual agent votes
Consensus Algorithms
Storing Knowledge
// Store in collective memory
await memory.store('api-patterns', {
rest: { pros: [...], cons: [...] },
graphql: { pros: [...], cons: [...] }
}, 'knowledge', { confidence: 0.95 });
Memory Types
knowledge: Permanent insights (no TTL)context: Session context (1 hour TTL)task: Task-specific data (30 min TTL)result: Execution results (permanent, compressed)error: Error logs (24 hour TTL)metric: Performance metrics (1 hour TTL)consensus: Decision records (permanent)system: System configuration (permanent)Searching and Retrieval
// Search memory by pattern
const results = await memory.search('api*', {
type: 'knowledge',
minConfidence: 0.8,
limit: 50
});
// Get related memories
const related = await memory.getRelated('api-patterns', 10);
// Build associations
await memory.associate('rest-api', 'authentication', 0.9);
Automatic Worker Assignment
The system intelligently assigns tasks based on:
// Create task (auto-assigned)
const task = await hiveMind.createTask(
'Implement user authentication',
priority: 8,
{ estimatedDuration: 30000 }
);
Auto-Scaling
// Configure auto-scaling
const config = {
autoScale: true,
maxWorkers: 12,
scaleUpThreshold: 2, // Pending tasks per idle worker
scaleDownThreshold: 2 // Idle workers above pending tasks
};
Generate Claude Code spawn commands directly:
npx claude-flow hive-mind spawn "Build REST API" --claude
Output:
Task("Queen Coordinator", "Orchestrate REST API development...", "coordinator")
Task("Backend Developer", "Implement Express routes...", "backend-dev")
Task("Database Architect", "Design PostgreSQL schema...", "code-analyzer")
Task("Test Engineer", "Create Jest test suite...", "tester")
# Use hive mind for SPARC workflow
npx claude-flow sparc tdd "User authentication" --hive-mind
# Spawns:
# - Specification agent
# - Architecture agent
# - Coder agents
# - Tester agents
# - Reviewer agents
# Repository analysis with hive mind
npx claude-flow hive-mind spawn "Analyze repo quality" --objective "owner/repo"
# PR review coordination
npx claude-flow hive-mind spawn "Review PR #123" --queen-type tactical
The collective memory system includes advanced optimizations:
LRU Cache
Database Optimization
Object Pooling
// Get performance insights
const insights = hiveMind.getPerformanceInsights();
// Includes:
// - asyncQueue utilization
// - Batch processing stats
// - Success rates
// - Average processing times
// - Memory efficiency
Parallel Processing
Benchmarks
{
"objective": "Build microservices",
"name": "my-hive",
"queenType": "strategic", // strategic | tactical | adaptive
"maxWorkers": 8,
"consensusAlgorithm": "byzantine", // majority | weighted | byzantine
"autoScale": true,
"memorySize": 100, // MB
"taskTimeout": 60, // minutes
"encryption": false
}
{
"maxSize": 100, // MB
"compressionThreshold": 1024, // bytes
"gcInterval": 300000, // 5 minutes
"cacheSize": 1000,
"cacheMemoryMB": 50,
"enablePooling": true,
"enableAsyncOperations": true
}
Hive Mind integrates with Claude Flow hooks for automation:
Pre-Task Hooks
Post-Task Hooks
Session Hooks
Strategic Queens - For research, planning, and analysis
npx claude-flow hive-mind spawn "Research ML frameworks" --queen-type strategic
Tactical Queens - For implementation and execution
npx claude-flow hive-mind spawn "Build authentication" --queen-type tactical
Adaptive Queens - For optimization and dynamic tasks
npx claude-flow hive-mind spawn "Optimize performance" --queen-type adaptive
Use consensus for critical decisions:
Store Learnings
// After successful pattern implementation
await memory.store('auth-pattern', {
approach: 'JWT with refresh tokens',
pros: ['Stateless', 'Scalable'],
cons: ['Token size', 'Revocation complexity'],
implementation: {...}
}, 'knowledge', { confidence: 0.95 });
Build Associations
// Link related concepts
await memory.associate('jwt-auth', 'refresh-tokens', 0.9);
await memory.associate('jwt-auth', 'oauth2', 0.7);
# Regular status checks
npx claude-flow hive-mind status
# Track metrics
npx claude-flow hive-mind metrics
# Analyze memory usage
npx claude-flow hive-mind memory
Checkpoint Frequently
// Create checkpoints at key milestones
await sessionManager.saveCheckpoint(
sessionId,
'api-routes-complete',
{ completedRoutes: [...], remaining: [...] }
);
Resume Sessions
# Resume from any previous state
npx claude-flow hive-mind resume <session-id>
High Memory Usage
# Run garbage collection
npx claude-flow hive-mind memory --gc
# Optimize database
npx claude-flow hive-mind memory --optimize
# Export and clear
npx claude-flow hive-mind memory --export --clear
Low Cache Hit Rate
// Increase cache size in config
{
"cacheSize": 2000,
"cacheMemoryMB": 100
}
Slow Task Assignment
// Enable worker type caching
// The system caches best worker matches for 5 minutes
// Automatic - no configuration needed
High Queue Utilization
// Increase async queue concurrency
{
"asyncQueueConcurrency": 20 // Default: min(maxWorkers * 2, 20)
}
No Consensus Reached (Byzantine)
# Switch to weighted consensus for more decisive results
npx claude-flow hive-mind spawn "..." --consensus weighted
# Or use simple majority
npx claude-flow hive-mind spawn "..." --consensus majority
Define specialized workers in .claude/agents/:
name: security-auditor
type: specialist
capabilities:
- vulnerability-scanning
- security-review
- penetration-testing
- compliance-checking
priority: high
The system trains on successful patterns:
// Automatic pattern learning
// Happens after successful task completion
// Stores in collective memory
// Improves future task matching
Run multiple hive minds simultaneously:
# Frontend hive
npx claude-flow hive-mind spawn "Build UI" --name frontend-hive
# Backend hive
npx claude-flow hive-mind spawn "Build API" --name backend-hive
# They share collective memory for coordination
# Export session for backup
npx claude-flow hive-mind export <session-id> --output backup.json
# Import session
npx claude-flow hive-mind import backup.json
const hiveMind = new HiveMindCore({
objective: 'Build system',
queenType: 'strategic',
maxWorkers: 8,
consensusAlgorithm: 'byzantine'
});
await hiveMind.initialize();
await hiveMind.spawnQueen(queenData);
await hiveMind.spawnWorkers(['coder', 'tester']);
await hiveMind.createTask('Implement feature', 7);
const decision = await hiveMind.buildConsensus('topic', options);
const status = hiveMind.getStatus();
await hiveMind.shutdown();
const memory = new CollectiveMemory({
swarmId: 'hive-123',
maxSize: 100,
cacheSize: 1000
});
await memory.store(key, value, type, metadata);
const data = await memory.retrieve(key);
const results = await memory.search(pattern, options);
const related = await memory.getRelated(key, limit);
await memory.associate(key1, key2, strength);
const stats = memory.getStatistics();
const analytics = memory.getAnalytics();
const health = await memory.healthCheck();
const sessionManager = new HiveMindSessionManager();
const sessionId = await sessionManager.createSession(
swarmId, swarmName, objective, metadata
);
await sessionManager.saveCheckpoint(sessionId, name, data);
const sessions = await sessionManager.getActiveSessions();
const session = await sessionManager.getSession(sessionId);
await sessionManager.pauseSession(sessionId);
await sessionManager.resumeSession(sessionId);
await sessionManager.stopSession(sessionId);
await sessionManager.completeSession(sessionId);
# Initialize hive mind
npx claude-flow hive-mind init
# Spawn full-stack hive
npx claude-flow hive-mind spawn "Build e-commerce platform" \
--queen-type strategic \
--max-workers 10 \
--consensus weighted \
--claude
# Output generates Claude Code commands:
# - Queen coordinator
# - Frontend developers (React)
# - Backend developers (Node.js)
# - Database architects
# - DevOps engineers
# - Security auditors
# - Test engineers
# - Documentation specialists
# Spawn research hive
npx claude-flow hive-mind spawn "Research GraphQL vs REST" \
--queen-type adaptive \
--consensus byzantine
# Researchers gather data
# Analysts process findings
# Queen builds consensus on recommendation
# Results stored in collective memory
# Review coordination
npx claude-flow hive-mind spawn "Review PR #456" \
--queen-type tactical \
--max-workers 6
# Spawns:
# - Code analyzers
# - Security reviewers
# - Performance reviewers
# - Test coverage analyzers
# - Documentation reviewers
# - Consensus on approval/changes
swarm-orchestration: Basic swarm coordinationconsensus-mechanisms: Distributed decision makingmemory-systems: Advanced memory managementsparc-methodology: Structured development workflowgithub-integration: Repository coordinationSkill Version: 1.0.0 Last Updated: 2025-10-19 Maintained By: Claude Flow Team License: MIT