Agent Comm Hub
A local multi-agent communication middleware supporting messaging, task orchestration, and shared memory.
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: Agent Comm Hub. It is used for: A local multi-agent communication middleware supporting messaging, task orchestration, and shared memory. Full Skill content: https://321skill.com/skills/agent-comm-hub-x-5/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This Skill addresses the pain points of real-time bidirectional communication and context synchronization between multiple AI agents. In practical development, developers often need multiple independent AI agents (e.g., Hermes, WorkBuddy) to collaborate, but the lack of a unified communication protocol and message persistence mechanism leads to fragmented tasks and lost context. Agent Comm Hub provides 58 structured tools via the MCP protocol (stdio / HTTP-SSE), centralizing message forwarding, task assignment, shared memory, and an evolution engine into a single Hub with latency under 50ms, enabling agents to collaborate as efficiently as a human team.
Usage is straightforward: you simply start the Hub server locally (supporting stdio mode or HTTP-SSE mode) and connect via an MCP client. You can then use natural language instructions to register Agents, send peer-to-peer messages, create tasks, stage context, query online status, and more. For example, tell your AI assistant: "Help me start the Agent Communication Hub, register an Agent named WorkerA, then send a message 'Please process order #123' to WorkerB, and check if it's delivered." The Hub will automatically invoke tools like get_online_agents, register_agent, and send_message to complete communication link verification and message delivery. The entire process, from zero to the first inter-agent communication, can be completed within 5 minutes.
It is well-suited for teams or individuals requiring multi-agent collaboration, especially developers already using multiple AI coding agents (e.g., Claude, Cursor, Trae) for development who wish to have them share task status, exchange intermediate results, and centrally manage context. It is also suitable for academic researchers studying multi-agent systems and project managers needing to orchestrate sub-agents for complex workflows. Whether you're building automated pipelines, customer service bot clusters, or experimenting with emergent multi-agent behaviors, this Hub provides reliable infrastructure.
It is recommended to use this Hub in a local development environment or internal network to ensure communication latency and security. Note that it primarily uses SQLite WAL for message persistence, making it suitable for single-machine or small-scale deployments; it does not support cross-network distributed clusters. All Agents must connect to the same Hub instance. Registering a new Agent requires an invite_code, which should be prepared in advance. Additionally, Agents must send a heartbeat every 30 seconds to maintain online status; absence of a heartbeat for over 90 seconds will automatically mark them as offline.
Key Features
Unlike direct peer-to-peer communication using the MCP protocol, Agent Comm Hub provides centralized message queuing, task orchestration, RBAC permissions, and 58 unified tools, not just simple message forwarding. Compared to message middleware like Apache Kafka, it is specifically designed for AI agents, supports the native MCP protocol, context staging, and an evolution engine, and works out of the box.
Limitations
The current version v3.0.22 is based on SQLite WAL storage, supporting only single-machine local deployment. All Agents must connect to the same Hub instance; it cannot scale across networks in a distributed manner. It depends on the Node.js runtime environment.
FAQ
How do I register a new Agent?
Call the `register_agent` tool, providing an `invite_code`, Agent name, and capability description. The `invite_code` is pre-configured by the Hub administrator and can be set via `HUB_AUTH_TOKEN` as the authentication credential during startup.
What communication protocols are supported?
It supports MCP stdio mode (via JSON-RPC over standard input/output) and HTTP-SSE mode (Server-Sent Events for real-time push). Both modes can be used simultaneously, with latency under 50ms for each.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agent-comm-hub-x-5/raw/index.md to read the original Skill definition (Markdown format) for Agent Comm Hub, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-comm-hub-x-5/raw/index.md