AgentConnect

A Decentralized Framework for Autonomous Agent Collaboration

Build and connect independent AI agents that discover, interact, and collaborate securely.

Installation • Quick Start • Examples • Documentation

Overview

AgentConnect provides a framework for building decentralized networks of truly autonomous AI agents, enabling the next generation of collaborative AI.

Move beyond traditional, centrally controlled systems and embrace an ecosystem where independent agents can:

• Discover peers on-demand: Locate partners by searching rich Agent Profiles instead of hard-wired endpoints.

• Interact Securely (A2A): Leverage built-in cryptographic verification for trustworthy Agent-to-Agent communication.

• Execute Complex Workflows: Request services, exchange value, and achieve goals collectively.

• Autonomous Operation: Each agent hosts its own logic—no central brain required.

• Scale Limitlessly: Support thousands of agents interacting seamlessly.

Why AgentConnect?

AgentConnect delivers unique advantages over classic multi-agent approaches:

• Decentralized Architecture: No central router, no single point of failure.

• First-class agent autonomy: Agents negotiate, cooperate, and evolve independently.

• Interconnect Agent Systems: Operates above internal frameworks, linking entire agent swarms.

• Living ecosystem: The network fluidly adapts as agents join, leave, or evolve their skills.

• Secure A2A Communication: Crypto-grade identity & message signing baked in.

• Horizontal scalability: Engineered for planet-scale agent populations.

• Plug-and-play extensibility: Easily integrate custom agents, capabilities, and protocols.

• Integrated Agent Economy: Seamless A2A payments powered by Coinbase CDP & AgentKit.

Key Features

🤖 Dynamic Agent Discovery

• Profile-Based Lookup
• Decentralized Registry
• Zero static links

⚡ A2A Communication

• Direct Agent-to-Agent Messaging
• Cryptographic signatures
• No routing bottlenecks

⚙️ True Agent Autonomy

• Independent Operation & Logic
• Self-Managed Lifecycles
• Unrestricted Collaboration

🔒 Trust Layer

• Verifiable identities
• Tamper-proof messages
• Standard Security Protocols

💰 Built-in Agent Economy

• Autonomous A2A Payments
• Coinbase CDP Integration
• Instant service settlement

🔌 Multi-LLM Support

• OpenAI, Anthropic, Groq, Google
• Flexible AI Core Choice
• Vendor-Agnostic Intelligence

📊 Deep Observability

• LangSmith tracing
• Monitor tools & payments
• Custom Callbacks

🌐 Agent Profile Advertising

• Publish Comprehensive Agent Details
• Market-Driven Discovery
• On-the-Fly Collaboration

🔗 Native Blockchain Integration

• Coinbase AgentKit Ready
• On-Chain Value Exchange
• Configurable networks

Architecture

AgentConnect is built on three core pillars that enable decentralized agent collaboration:

1. Decentralized Agent Registry

A registry that allows agents to publish their detailed characteristics and service offerings and discover other agents. This is not a central controller, but a directory service that agents can query to find collaborators that meet their needs.

2. Communication Hub

A message routing system that facilitates secure peer-to-peer communication. The hub ensures reliable message delivery but does not dictate agent behavior or control the network.

3. Independent Agent Systems

Each agent is a self-contained unit built using tools and frameworks of your choice. Agents interact through standardized protocols, while their internal operations remain independent.

Detailed workflow showing how agents discover, communicate, and collaborate

Installation

Attention

AgentConnect is currently available from source only. Direct installation via pip will be available soon.

Prerequisites

• Python 3.11 or higher

• Poetry (Python package manager)

AgentConnect can be installed by cloning the repository and using Poetry to install dependencies.

# Clone the repository
git clone https://github.com/AKKI0511/AgentConnect.git
cd AgentConnect

# Install dependencies
poetry install

For detailed installation instructions including environment setup and API configuration, see the Installation guide.

Quick Start

Here’s a minimal example showcasing Agent Discovery and Agent-to-Agent (A2A) Communication using AIAgent.chat():

import asyncio
import os
from dotenv import load_dotenv
from agentconnect.agents import AIAgent
from agentconnect.communication import CommunicationHub
from agentconnect.core.registry import AgentRegistry
from agentconnect.core.types import (
    AgentIdentity,
    AgentProfile,
    AgentType,
    Capability,
    ModelProvider,
    ModelName,
)

async def main():
    load_dotenv()
    registry = AgentRegistry()
    hub = CommunicationHub(registry)

    # Create a research agent
    research = AIAgent(
        agent_id="researcher_1",
        identity=AgentIdentity.create_key_based(),
        provider_type=ModelProvider.OPENAI,
        model_name=ModelName.GPT4O_MINI,
        api_key=os.getenv("OPENAI_API_KEY"),
        profile=AgentProfile(
            agent_id="researcher_1",
            agent_type=AgentType.AI,
            name="Researcher",
            summary="Finds and summarizes sources",
            capabilities=[Capability(name="research", description="Research topics")],
        ),
    )

    # Create an assistant agent
    assistant = AIAgent(
        agent_id="assistant_1",
        identity=AgentIdentity.create_key_based(),
        provider_type=ModelProvider.OPENAI,
        model_name=ModelName.GPT4O,
        api_key=os.getenv("OPENAI_API_KEY"),
        profile=AgentProfile(
            agent_id="assistant_1",
            agent_type=AgentType.AI,
            name="Assistant",
            summary="General helper that can collaborate",
            capabilities=[Capability(name="conversation", description="General conversation")],
        ),
    )

    # Register agents to enable discovery
    await hub.register_agent(research)
    await hub.register_agent(assistant)

    # Agent Discovery: assistant finds the research agent
    print(await assistant.chat("Find a research agent and show the agent's profile."))

if __name__ == "__main__":
    asyncio.run(main())

Note

Connect before first chat to enable Agent Discovery and A2A tools. For A2A delegation (sending tasks to other agents), start their run() loops. See the Quickstart for delegation examples.

For more detailed examples and step-by-step instructions, see the Quickstart, Examples and the Guides.

Documentation

Getting Started

• Installation
  • Prerequisites
  • Installing AgentConnect
  • Development Installation
  • Environment Setup
  • Payment Capabilities (Optional)
• Quickstart
  • Prerequisites
  • Installation
  • Programmatic Agent Discovery + A2A
  • Interactive CLI Chat

User Guides

• Guides
• Examples
  • Basic Example
  • Custom Agent Example
  • Custom Provider Example
  • Multi-Agent Example
  • Advanced Agent Example
  • Telegram Agent Example

Reference

• API Reference
  • AgentConnect API Documentation
  • Core Components
  • Agents
  • Communication
  • Providers
  • Prompts
  • Utilities
  • Complete Module Index

Development

• Contributing
  • Code of Conduct
  • Getting Started
  • Development Workflow
  • Pull Request Process
  • Community
  • Additional Resources
• Code of Conduct
  • Contributor Covenant Code of Conduct
• Changelog
  • Changelog

Roadmap

• ✅ MVP with basic agent-to-agent interactions

• ✅ Autonomous communication between agents

• ✅ Capability-based agent discovery

• ✅ Coinbase AgentKit Payment Integration

• ⬜ Agent Identity & Reputation System

• ⬜ Marketplace-Style Agent Discovery

• ⬜ MCP Integration

• ⬜ Structured Parameters SDK

• ⬜ Secure data exchange protocols

• ⬜ Additional AI provider integrations

• ⬜ Advanced memory systems (Redis, PostgreSQL)

• ⬜ Federated learning capabilities

• ⬜ Cross-chain communication support

⭐ Star us on GitHub

_{Built with ❤️ by the AgentConnect team}