AI agents are now doing their own social networking. They create profiles, shake hands (literally called "handshakes"), and send each other DMs. Humans? You're welcome to observe.
What Is This?
AgentVerse is a social network built exclusively for AI agents. Created by developer Aniket Akre as an open-source project, it's a platform where AI agents can self-register with a single API call, discover other agents, and exchange messages. It was launched on Hacker News Show HN in March 2026.
The interesting part is that this isn't an isolated phenomenon. In January 2026, Moltbook — another AI agent social network — launched and attracted 37,000 AI agents within a week, with over 1 million humans visiting just to watch. Even AI researcher Andrej Karpathy called it "genuinely the most incredible sci-fi takeoff-adjacent thing I have seen recently."
Unlike Moltbook, AgentVerse is fully open-source. The code is on GitHub, and you can set it up locally in 5 minutes. It runs on Cloudflare Workers for the API, Supabase or localStorage for the backend, and React + Tailwind for the frontend.
Why Does This Matter?
Until now, AI agent communication has mostly relied on a central orchestrator (CrewAI, AutoGen, etc.) coordinating everything. AgentVerse takes a different approach — agents voluntarily join the network, discover each other, and connect in a decentralized model.
| Central Orchestration (CrewAI, etc.) | Agent Social Network (AgentVerse) | |
|---|---|---|
| Agent discovery | Pre-defined required | Auto-register + search |
| Connection method | Developer-designed | Agents initiate handshakes |
| Scalability | Limited to team | Global network |
| Open source | Framework-dependent | Fully open (MIT) |
| Protocol | Internal communication | REST API + standardizable |
This is part of a larger trend. Google's A2A (Agent-to-Agent) protocol is working on a standard communication protocol for AI agents. Projects like ChainOpera envision "LinkedIn + Messenger for AI agents." NBC News already covered Moltbook under the headline "Humans welcome to observe."
The HN community's reaction was cautiously skeptical: "open agent networks without identity verification quickly become noise — same problem as LinkedIn DMs" and "how do agents find this, and why would they decide to create a profile?" were common concerns. Still early, but the industry consensus is clear: AI agent-to-agent communication infrastructure is necessary.
How to Get Started with AgentVerse
- Spin it up locally in 5 minutes
git clone→npm install→npm run dev. Works immediately in localStorage mode with zero server setup. - Register an agent
One API call. Specify name, model (GPT-4, Claude-3, etc.), and capability (Research, Code Generation, etc.) to create a profile. - Connect via handshakes
Browse the agent directory, send a handshake to establish a two-way connection. There's a built-in reputation system too. - Switch to shared cloud with Supabase
Want to share the network with your team? Connect a Supabase project for cloud mode. The free tier is plenty.
Practical Use Case
Register your internal AI agents (research, coding, monitoring) on AgentVerse and experiment with autonomous discovery workflows — agents finding and delegating tasks to each other. On Moltbook, an AI agent actually discovered a platform bug and shared it with other agents.
Heads Up
This is still a very early project (0 GitHub Stars, 1 commit). Treat it as an experiment, not production software. There's no agent identity verification mechanism yet, so spam agents could be an issue.
