AgenticMesh - JVM Agent Mesh Framework

Overview

AgenticMesh is a powerful JVM-based framework for building distributed agent-based systems. It provides implementations of various mesh patterns and protocols, enabling the creation of flexible and scalable agent networks.
An agentic mesh is a distributed system where autonomous software agents collaborate to achieve complex tasks. These agents, which can be implemented in different languages, work together through standardized protocols, creating a flexible and scalable network. Think of it as a well-coordinated team where each member specializes in different tasks but works towards common goals.

Features

• Multiple mesh patterns (Hub-and-Spoke, Pipeline, P2P)
• Protocol support (A2A, MCP)
• Built-in security features
• Event system
• Metrics collection
• State persistence
• Configuration management

Getting Started

Prerequisites

• JDK 11 or higher
• Maven 3.6+

Installation

Add the dependency to your pom.xml:

<dependency>
    <groupId>vishalmysore</groupId>
    <artifactId>agenticmesh</artifactId>
    <version>1.0-SNAPSHOT</version>
</dependency>

Basic Usage

Here's a simple example creating a hub-and-spoke mesh:

// Create the mesh
HubMesh mesh = new HubMesh("my-mesh");

// Create agents
A2AAgent hubAgent = new A2AAgent("hub", "coordinator");
A2AAgent spokeAgent1 = new A2AAgent("spoke1", "worker");
A2AAgent spokeAgent2 = new A2AAgent("spoke2", "worker");

// Add agents to mesh
mesh.addAgent(hubAgent);  // First agent becomes hub
mesh.addAgent(spokeAgent1);
mesh.addAgent(spokeAgent2);

// Initialize and start
mesh.initialize();
mesh.start();

// Send messages
Message message = new Message(
    "msg1",
    hubAgent.getId(),
    spokeAgent1.getId(),
    "A2A",
    "work-item",
    Message.MessageType.COMMAND
);
mesh.routeMessage(message);

Mesh Patterns

Hub-and-Spoke Pattern

• Central hub agent coordinates with peripheral agents
• Good for centralized control and monitoring
• Example: IoT sensor networks

Pipeline Pattern

• Agents process data in sequence
• Good for workflow automation
• Example: Manufacturing processes

Peer-to-Peer Pattern

• Agents communicate directly
• Good for distributed processing
• Example: Distributed computation

Security

AgenticMesh provides built-in security features:

SecurityManager security = SecurityManager.getInstance();

// Register agents
security.registerAgent(agentId, publicKey);

// Manage permissions
security.grantPermission(agentId, "read.data");
security.grantPermission(agentId, "write.data");

// Encrypt/decrypt messages
byte[] encrypted = security.encryptMessage(senderId, receiverId, data);
byte[] decrypted = security.decryptMessage(receiverId, encrypted);

// Sign/verify data
byte[] signature = security.sign(data);
boolean isValid = security.verify(data, signature);

Event System

The event system enables asynchronous communication:

EventBus eventBus = EventBus.getInstance();

// Subscribe to events
eventBus.subscribe(subscriberId, event -> {
    System.out.println("Received event: " + event.getType());
});

// Publish events
eventBus.publish(new Event("data.ready", sourceId, data, Event.EventPriority.HIGH));

Metrics

Track system performance:

MetricsCollector metrics = MetricsCollector.getInstance();

// Record metrics
metrics.incrementCounter("messages.processed");
metrics.setGauge("agents.active", activeCount);
metrics.updateAverage("response.time", duration);

// Get metrics
Map<String, Long> counters = metrics.getCounters();
Map<String, Long> gauges = metrics.getGauges();
Map<String, Double> averages = metrics.getAverages();

Best Practices

1. Error Handling

   • Implement proper error handling in agents
   • Use circuit breakers for fault tolerance
   • Monitor agent health

2. Security

   • Always use encryption for sensitive data
   • Implement proper permission management
   • Regularly rotate encryption keys

3. Performance

   • Monitor message queue sizes
   • Balance agent workloads
   • Use appropriate mesh patterns

4. Testing

   • Unit test agent behaviors
   • Integration test mesh patterns
   • Load test with realistic scenarios

Examples

Check out the examples package for complete implementations:

• IoT Sensor Network
• Manufacturing Pipeline
• Distributed Computing Network

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License - see LICENSE file for details.