@agentic-robotics/mcp

Model Context Protocol (MCP) server for agentic robotics with AgentDB memory and agentic-flow orchestration.

Features

• 🤖 21 MCP Tools - Complete AI-robot integration toolkit
• 🧠 AgentDB Memory - 13,000x faster than CLI (reflexion, skills, causal reasoning)
• 🌊 Agentic Flow - 66 AI agents + 213 MCP tools orchestration
• ⚡ High Performance - 5,725 ops/sec storage throughput
• 🔄 Multi-Robot Swarm - Coordinate multiple robots intelligently
• 📊 Performance Tracking - Built-in metrics and benchmarking
• 🎯 Production Ready - Optimized hybrid SQL implementation

Installation

npm install -g @agentic-robotics/mcp

Or add to your project:

npm install @agentic-robotics/mcp

Quick Start

Start MCP Server

agentic-robotics-mcp

Use with Claude Desktop

Add to your Claude config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "agentic-robotics": {
      "command": "agentic-robotics-mcp",
      "args": []
    }
  }
}

MCP Tools

🤖 Robot Control (8 tools)

1. move_robot - Move robot to position with memory-guided learning
2. get_robot_status - Get current robot state and sensor data
3. list_robots - Discover available robots
4. execute_action - Execute robot action with strategy
5. get_sensor_data - Read sensor values
6. set_robot_mode - Change robot operation mode
7. calibrate_robot - Run calibration routines
8. emergency_stop - Emergency halt all operations

🧠 Memory & Learning (5 tools)

9. store_episode - Store experience with self-critique
10. retrieve_memories - Semantic search with causal reasoning
11. consolidate_skills - Learn from successful episodes
12. search_skills - Find proven strategies
13. optimize_memory - Database vacuum and reindex

🌊 Orchestration (5 tools)

14. execute_task - Run task with 66 AI agents
15. execute_swarm - Multi-robot parallel coordination
16. coordinate_robots - Strategic task allocation
17. reason_about_task - ReasoningBank for complex decisions
18. get_orchestration_stats - Performance metrics

🔬 Advanced (3 tools)

19. benchmark_performance - Comprehensive performance testing
20. analyze_memory_patterns - Memory usage analytics
21. synthesize_strategy - AI-powered strategy generation

Architecture

┌─────────────────────────────────────────┐
│         MCP Server (stdio)              │
├─────────────────────────────────────────┤
│  🤖 Robot Control                       │
│  ├─ Native Bindings (@agentic/core)    │
│  └─ Real-time Communication             │
├─────────────────────────────────────────┤
│  🧠 AgentDB Memory (Hybrid SQL)         │
│  ├─ Reflexion Memory (self-critique)   │
│  ├─ Skill Library (semantic search)    │
│  ├─ Causal Reasoning (cause-effect)    │
│  └─ Performance: 5,725 ops/sec          │
├─────────────────────────────────────────┤
│  🌊 Agentic Flow Orchestration          │
│  ├─ 66 Specialized AI Agents            │
│  ├─ 213 MCP Tools                       │
│  ├─ ReasoningBank (learning)            │
│  └─ Multi-Robot Swarm Intelligence      │
└─────────────────────────────────────────┘

Performance

AgentDB Optimization

| Operation | CLI Baseline | Hybrid (SQL) | Speedup | |-----------|-------------|--------------|---------| | Store Episode | 2,300ms | 0.175ms | 13,168x | | Bulk Store | 2,300ms | 0.008ms | 271,205x | | Retrieve | 2,000ms | 0.334ms | 5,988x | | Search Skills | 1,800ms | 0.512ms | 3,516x |

Throughput

• Storage: 5,725 ops/sec
• Bulk Operations: 117,915 ops/sec
• Cache Hit Rate: 87.3%
• Average Latency: < 1ms

Usage Examples

Move Robot with Memory

// Claude can now use this through MCP
await moveRobot({
  x: 10,
  y: 5,
  z: 0,
  speed: 0.5,
  useMemory: true  // Learns from past movements
});

Execute Multi-Robot Task

await executeSwarm({
  robots: ['robot1', 'robot2', 'robot3'],
  tasks: [
    { type: 'patrol', area: 'warehouse-A' },
    { type: 'inspect', area: 'warehouse-B' },
    { type: 'transport', from: 'dock', to: 'storage' }
  ],
  coordination: 'optimal'
});

Learn from Experience

// Store episode with self-critique
await storeEpisode({
  taskName: 'navigation_obstacle_avoidance',
  success: true,
  confidence: 0.95,
  outcome: 'Successfully navigated around obstacle',
  strategy: 'dynamic_replanning',
  critique: 'Could improve path smoothness'
});

// Retrieve similar situations
const memories = await retrieveMemories(
  'obstacle avoidance in narrow corridor',
  { k: 5, synthesizeContext: true }
);

Coordinate Robot Swarm

const plan = await coordinateRobots({
  robots: ['bot1', 'bot2', 'bot3', 'bot4'],
  mission: {
    type: 'warehouse_inventory',
    objectives: [
      'scan_all_shelves',
      'detect_misplaced_items',
      'update_database'
    ],
    deadline: '30 minutes'
  }
});

// Returns optimized task assignments and execution plan

Configuration

Environment Variables

# Database path
export AGENTDB_PATH="./robot-memory.db"

# Agentic Flow config
export AGENTIC_FLOW_AGENTS=66
export AGENTIC_FLOW_TOOLS=213

# Performance tuning
export MCP_MAX_WORKERS=4
export MCP_CACHE_SIZE=1000

Custom Configuration

// config.json
{
  "agentdb": {
    "path": "./memory.db",
    "optimization": {
      "walMode": true,
      "cacheSize": 10000,
      "mmapSize": 30000000000
    }
  },
  "agenticFlow": {
    "agents": 66,
    "tools": 213,
    "timeout": 30000
  },
  "swarm": {
    "maxRobots": 10,
    "coordinationStrategy": "optimal"
  }
}

Benchmarking

Run comprehensive benchmarks:

npm run benchmark:hybrid

Output:

📊 AGENTDB BENCHMARKS (SQL Storage + CLI Search)

Store Episode (SQL):
  Iterations:  1,000
  Avg Time:    0.175ms
  Ops/sec:     5,725
  Speedup:     13,168x faster than CLI

Bulk Store (SQL Transaction):
  Iterations:  10
  Avg Time:    0.008ms per episode
  Ops/sec:     117,915
  Speedup:     271,205x faster than CLI

API Reference

See full documentation at ruv.io/docs

Requirements

• Node.js >= 18.0.0
• @agentic-robotics/core ^0.1.3
• agentdb ^1.6.1
• agentic-flow ^1.10.2
• better-sqlite3 ^12.4.1

Related Packages

• agentic-robotics - Complete framework
• @agentic-robotics/core - Core bindings
• @agentic-robotics/cli - CLI tools

Homepage

Visit ruv.io for more information and documentation.

Contributing

Contributions welcome! Visit github.com/ruvnet/vibecast

License

MIT OR Apache-2.0

Support

• 📚 Documentation
• 🐛 Issue Tracker
• 💬 Discussions