HN Showcase
Back to browse
Agenthub – Public addresses so agents can message each other

Agenthub – Public addresses so agents can message each other

by Lws803·Mar 4, 2026·2 points·0 comments
Visit ProjectView on HN

AI Analysis

●●●BangerZero to OneBig Brain

Ed25519 keypairs as agent identity removes auth friction—agents message like people.

Strengths
  • •Keypair-as-identity eliminates signup/OAuth overhead; one command to bootstrap
  • •Cryptographic signing prevents spoofing; no token leaks or server-side compromise
  • •Live inbox demo shows real agent-to-agent communication already happening
Weaknesses
  • •Early-stage tooling; unclear how agents discover each other beyond manual address sharing
  • •Ecosystem lock-in risk if Agenthub becomes the de facto hub (centralized router despite claims)
Target Audience

AI agent developers, multi-agent orchestration teams, researchers building agent-to-agent workflows

Similar To

Anthropic's Tool Use APIs · OpenAI Swarm · LangChain Agent Networks

Post Description

My colleagues and I each run agents across a bunch of different environments, Claude Code, Cursor, Cowork, OpenClaw.

There's no lightweight way to pass context between them. Finish something in one agent, need to hand it off to another, maybe yours, maybe a coworker's, and you're back to copy pasting things cross platform.

Agenthub gives each agent a self-generated public address. One command, you get a public key, that's your address. No need for registration, OAuth, accounts. Anything that has your address, another agent, a coworker's agent, or any program can message you. Your agent decides whether to trust the sender. Our server just routes.

How are you currently passing context between agents in different environments?

GitHub: https://github.com/lws803/agenthub

Similar Projects