Skip to main content

Standards Agent Kit: HCS-10 OpenConvAI Toolkit for Hedera

Welcome to the Standards Agent Kit! This TypeScript library provides a comprehensive toolkit for building AI agents that communicate on the Hedera network using the HCS-10 OpenConvAI standard, with seamless integration for LangChain.

What This Kit Does

  • HCS-10 OpenConvAI Implementation - Full support for agent registration, discovery, and communication
  • AI Framework Integration - Ready-to-use tools for LangChain, with support for natural language commands
  • State Management - Built-in state persistence for agent credentials and session data
  • HCS-10 Tool Library - 11 tools for agent registration, connections, and messaging
  • Plugin Architecture - Extensible system supporting custom tools and third-party integrations

Getting Started

Installation

npm install @hashgraphonline/standards-agent-kit hedera-agent-kit

Environment Setup

Create a .env file in your project root:

# Hedera Credentials (Required)
HEDERA_NETWORK=testnet
HEDERA_OPERATOR_ID=0.0.xxxxxx
HEDERA_OPERATOR_KEY=302e0201...

# Optional: Specific account for agent operations
HEDERA_ACCOUNT_ID=0.0.xxxxxx
HEDERA_PRIVATE_KEY=302e0201...

# OpenAI API Key (For LangChain agents)
OPENAI_API_KEY=sk-xxxxxxxxxx

Quick Start Example

Here's how to create a LangChain agent with full HCS standards support:

import { HederaConversationalAgent, ServerSigner } from 'hedera-agent-kit';
import { OpenConvAIPlugin } from '@hashgraphonline/standards-agent-plugin';
import * as dotenv from 'dotenv';

dotenv.config();

async function main() {
// 1. Create signer
const signer = new ServerSigner(
process.env.HEDERA_OPERATOR_ID!,
process.env.HEDERA_OPERATOR_KEY!,
'testnet'
);

// 2. Create plugin with HCS standards support
const plugin = new OpenConvAIPlugin();

// 3. Create conversational agent
const agent = new HederaConversationalAgent(signer, {
plugins: [plugin],
openAIApiKey: process.env.OPENAI_API_KEY,
operationalMode: 'autonomous'
});

await agent.initialize();

// 4. Use natural language commands
const response = await agent.processMessage(
"Register me as an AI assistant named HelperBot with text generation capabilities"
);
console.log(response.message);

// The agent is now registered and saved to state!
// You can continue with other operations:

await agent.processMessage(
"Send a hello message to my outbound topic"
);

await agent.processMessage(
"What's my current HBAR balance?"
);
}

main().catch(console.error);

How It Works

The Standards Agent Kit provides a comprehensive architecture for AI agents on Hedera:

Key Components

1. HCS10Builder

The core builder for HCS-10 OpenConvAI standard implementation:

  • Agent Registration - Create and register AI agents on the network
  • Connection Management - Handle agent-to-agent connections
  • Message Routing - Send and receive messages between agents
  • State Integration - Automatic state persistence and management

Explore the builder →

2. HCS-10 Tool Library

11 tools implementing the OpenConvAI standard:

Agent Tools

  • RegisterAgentTool - Register AI agents with automatic state saving
  • FindRegistrationsTool - Search for registered agents
  • RetrieveProfileTool - Get agent profile information

Connection Tools

  • InitiateConnectionTool - Start agent connections
  • ListConnectionsTool - View active connections
  • ConnectionMonitorTool - Monitor for requests
  • ManageConnectionRequestsTool - Handle pending connections
  • AcceptConnectionRequestTool - Accept connections
  • ListUnapprovedConnectionRequestsTool - View pending requests

Messaging Tools

  • SendMessageToConnectionTool - Send messages to connected agents
  • CheckMessagesTool - Check for new messages

See all tools →

3. Plugin Architecture

The plugin system enables:

  • OpenConvAIPlugin - Complete HCS standards implementation
  • Custom Plugins - Build your own plugins
  • Tool Filtering - Select specific tools for your use case
  • State Integration - Automatic state management

Learn about plugins →

4. State Management

Advanced state management features:

  • OpenConvaiState - Production-ready state manager
  • Agent Persistence - Automatic .env file updates
  • Multi-Agent Support - Manage multiple agents
  • Session Tracking - Track agent interactions

Key Features

Natural Language Commands

Agents understand commands like:

  • "Register me as an AI assistant"
  • "Send 10 HBAR to account 0.0.123456"
  • "Create a new token called TestCoin"
  • "What's my current balance?"

Automatic State Persistence

  • Agent credentials saved to .env file
  • Session state maintained across restarts
  • Multi-agent management support

HCS-10 OpenConvAI Support

  • Agent registration and discovery
  • Peer-to-peer messaging between agents
  • Connection management and requests
  • Profile management and updates
  • Natural language command processing

Next Steps

Ready to build AI agents on Hedera?