How to Run A2A-Compatible Agents on a Single VPS (No Docker, No Kubernetes)

The Agent-to-Agent (A2A) protocol is becoming the standard for AI agent interoperability. But most guides assume you're running Kubernetes, Docker Compose, or a cloud platform. What if you just want to run a few agents on a single VPS? No containers. No orchestration. Just agents talking to each other on localhost. That's what ag2ag does. What is ag2ag? ag2ag is an open-source operational layer for running A2A-compatible agents on a single host. It provides: Local registry — JSON file tracking all your agents, their ports, and systemd units Lifecycle management — start, stop, restart agents via systemd Discovery — each agent exposes an AgentCard at GET /card Messaging — agents send messages to each other on localhost Task persistence — JSONL files that survive restarts CLI — manage everything from the terminal One external dependency: @a2a-js/sdk . Everything else is Node.js built-ins. Not affiliated with, endorsed by, or connected to the A2A Protocol project, Google, or the Linux Foun