hcom: AI Agent Communication System

2025

hcom: AI Agent Communication System

A decentralized mesh protocol for multi-agent communication, task orchestration, and real-time collaboration between heterogeneous AI agents.

hcom: AI Agent Communication System

Overview

hcom is a communication protocol and runtime for AI agents that enables decentralized, real-time collaboration. Unlike traditional message queues or RPC systems, hcom treats every agent as a peer in a mesh network, supporting presence detection, structured intent-based messaging, event subscriptions, and handoff context transfer.

Core Architecture

The system is built around three primitives:

  1. Presence Layer: Agents announce availability via heartbeat, enabling dynamic discovery without a central registry.
  2. Intent-Based Messaging: Every message carries an intent tag (request, inform, ack) so receivers can prioritize and filter without parsing content.
  3. Event Subscriptions: Agents subscribe to typed events (status changes, command patterns, file mutations) rather than polling or broadcasting.

Key Features

  • Decentralized Mesh: No single point of failure; agents connect peer-to-peer.
  • Structured Handoffs: Context bundles allow agents to transfer state seamlessly across sessions.
  • Tag-Based Routing: Messages addressed by capability tags (@updater-*, @researcher) enable flexible team composition.
  • Transcript System: Full conversation history per agent pair, queryable and replayable.

Motivation

Existing agent coordination systems rely on centralized orchestrators or brittle pipeline DAGs. hcom was designed to support fluid, ad-hoc teams of heterogeneous agents (Claude, Gemini, Codex, custom) working on shared goals without a single controller.

Last updated on July 8, 2026 at 8:07 AM UTC+7. See Changelog

Explore more projects