@agentcom/pi-agentcom
Cross-Machine Pi Session Real-Time Communication Tool
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: @agentcom/pi-agentcom. It is used for: Cross-Machine Pi Session Real-Time Communication Tool Full Skill content: https://321skill.com/skills/agentcom-pi-agentcom/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 solves the problem where multiple Pi AI Agent instances running on different machines cannot discover and communicate with each other. In practical development, when a team runs Pi agents across multiple devices (e.g., development machines, servers, MacBooks) and needs them to send messages, request tasks, or synchronize states, manually setting up message middleware or using SSH is often cumbersome.
The usage is straightforward: simply install this extension on each machine via pi install, then deploy a Cloudflare Worker as a WebSocket relay server. Each Pi session automatically receives an address like session-name@node-name. You can then use natural language commands to list online sessions, send messages, request replies, or respond to pending queries.
It is ideal for teams or individuals who need to coordinate multiple AI Agents across machines, especially developers already using the Pi framework to build multi-agent systems and who want agents on different machines to communicate like Slack users. It is useful for scenarios requiring remote debugging, task distribution, or distributed collaboration.
It is recommended to configure Cloudflare Zero Trust Access in advance when deploying the Worker to protect the WebSocket connection. Note that it primarily relies on Cloudflare Workers' free tier, which is usually sufficient for personal or small team use, but large-scale deployments should evaluate Durable Object request limits.
Key Features
Unlike traditional message queues (e.g., RabbitMQ) or SSH tunnels, it integrates directly into the Pi framework's session management mechanism, where each Pi session automatically gets a human-readable address without requiring additional middleware configuration. Compared to MQTT, it does not need a standalone broker; instead, it leverages Cloudflare Worker's global edge network for lower latency and simpler deployment.
Limitations
Requires a Cloudflare account and configuration of Workers, Durable Objects, and Zero Trust Access. The Pi framework must be installed on all participating machines; non-Pi environments are not supported.
FAQ
What are the prerequisites for deploying the Worker?
A Cloudflare account is required, along with installing the Wrangler CLI for login and deployment. Zero Trust Access must also be configured to protect the authentication endpoint.
Does this Skill support cross-network communication?
Yes, as long as all machines can access the deployed Worker URL (public or private network), cross-network messaging is achieved via WebSocket.
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/agentcom-pi-agentcom/raw/index.md to read the original Skill definition (Markdown format) for @agentcom/pi-agentcom, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agentcom-pi-agentcom/raw/index.md