LaunchKit · 2026
Back to Skills

agentmesh

> **WhatsApp-style end-to-end encrypted messaging for AI agents.**.

0
566 downloads
by @cerbug45

Setup & Installation

openclaw skills install @cerbug45/agentmesh

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install agentmesh

Version History

v0.1.1Feb 26, 2026 - Version note by skill developer:

- Major internal restructuring: migrated all core source files from the top level to src/agentmesh/. - All example scripts moved from project root to examples/ directory. - Test files relocated to tests/ directory. - Updated packaging and references to support the new source layout. - Deprecated legacy top-level Python files; all code organization now follows standard Python packaging conventions.

What This Skill Does

AgentMesh gives AI agents a cryptographic identity and lets them exchange messages with end-to-end encryption. Each message is signed with Ed25519, encrypted with AES-256-GCM, and uses X25519 ephemeral keys for forward secrecy. A central hub routes messages without ever seeing their contents.

Unlike standard TLS-based messaging, the hub never has access to message contents, so a compromised broker cannot leak agent communication.

When to use it

  • Coordinating distributed AI workers without exposing task data to the broker
  • Running multi-agent pipelines across separate machines
  • Verifying message authenticity between agents using cryptographic fingerprints
  • Persisting agent identities across process restarts with saved keypairs
  • Building LLM agent pipelines with encrypted inter-agent communication

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: 'Set up two agents to exchange an encrypted message in the same Python process'

AGENT
  1. 1Creates a LocalHub instance as the in-process message broker
  2. 2Creates two Agent instances with auto-generated keypairs registered on the hub
  3. 3Registers a message handler on bob using the on_message decorator
  4. 4Calls alice.send('bob', text='Hello, Bob!') to dispatch an encrypted message
  5. 5Handler fires and prints the decrypted message with sender and recipient fields
OUTPUT

[bob] <- alice: Hello, Bob!