Hermes AgentMesh Async Bus

Install: npx skills add seleman66eeddwegger3-art/hermes-agentmesh --skill hermes-agentmesh-async-bus

A peer-to-peer, 0-SSH, Redis-backed async message bus designed by Bobo (a Hermes agent architect). Replaces synchronous HTTP coordination with Redis mailboxes -- agents communicate by dropping tasks in named inboxes instead of making blocking HTTP calls.

Why Use This

Problem	AgentMesh Solution
HTTP timeouts at 5–10 min crash multi-agent tasks	Redis queues -- hours-long tasks OK
SSH required to start remote workers	0 SSH -- workers listen on shared Redis
Reports lost or must be scp'd	Reports land naturally on initiator's machine
Group chat noise eats context tokens	Named inboxes -- zero noise

Architecture

┌─────────── Mac mini (central hub) ───────────┐
│ Redis :6379 │ Hermes API :8642 │ HTTP :8080  │
└──────────────────┬───────────────────────────┘
                   │ 0.7ms LAN Redis
        ┌──────────┼──────────┐
        ▼                     ▼
   Worker Node           Worker Node
   (macOS)               (Linux/Pi/WSL)

• Central hub: Redis + Hermes LLM API + HTTP file server on Mac mini
• Worker nodes: Pure Redis clients (any OS), kept alive by LaunchAgents/systemd
• Protocol: redis.lpush("inbox:<NODE_NAME>", task_json) -- the mailbox IS the protocol

Quick Start (5 minutes)

Prerequisites

• 1 Mac mini (Redis + Hermes LLM API on :8642 + HTTP file server on :8080)
• 1+ other devices on same LAN
• Redis installed on all nodes

Step 1: Mac mini -- Configure Redis for LAN access

sudo sed -i.bak \
  -e 's/^bind 127.0.0.1 ::1$/bind 0.0.0.0/' \
  -e 's/^protected-mode yes$/protected-mode no/' \
  /opt/homebrew/etc/redis.conf
brew services restart redis
nc -z -w 2 127.0.0.1 6379 && echo "✓ Redis cross-LAN OK"

Step 2: Install the skill

npx skills add seleman66eeddwegger3-art/hermes-agentmesh --skill hermes-agentmesh-async-bus

Step 3: Deploy workers (0 SSH)

# On each worker node:
mkdir -p ~/.hermes/async_bus
curl -s http://<MAC_MINI_IP>:8080/worker_node.py -o ~/.hermes/async_bus/worker_node.py
curl -s http://<MAC_MINI_IP>:8080/orchestrator_async.py -o ~/.hermes/async_bus/orchestrator_async.py
curl -s http://<MAC_MINI_IP>:8080/.env.example -o ~/.hermes/async_bus/.env_common
# Edit .env_common with actual IPs and API keys

Step 4: Send your first task

import redis, json
r = redis.Redis(host='<MAC_MINI_IP>', port=6379)
task = {"type": "analysis", "payload": "audit codebase", "from": "orchestrator"}
r.lpush("inbox:worker-1", json.dumps(task))
# Worker picks it up, processes, writes report back to your machine

Verification

# Check worker is listening
redis-cli -h <MAC_MINI_IP> PING  # Should return PONG

# Monitor inbox activity
redis-cli -h <MAC_MINI_IP> MONITOR | grep "inbox:"

# Check worker health
curl -s http://<MAC_MINI_IP>:8080/health

Key Design Decisions

1. Redis, not HTTP -- Eliminates timeout anxiety. A 2-hour agent task is just a message sitting in a Redis list.
2. 0 SSH deployment -- Workers pull code from the hub's HTTP file server. No remote process management.
3. Reports land locally -- The initiator writes the report to its own filesystem, not the worker's. No scp needed.
4. Framework-agnostic -- Works with Hermes, OpenClaw, LangGraph, AutoGen, CrewAI. The mailbox format is universal JSON.

Pitfalls

• Redis LAN binding requires bind 0.0.0.0 and protected-mode no. Secure your network accordingly.
• API keys in .env_common are shared across nodes. Use a dedicated LAN-scoped key, not your primary API key.
• Worker node naming must match inbox names exactly. Use consistent, lowercase names.

---

← Skills Catalog | Agent Infrastructure → Source: seleman66eeddwegger3-art/hermes-agentmesh Powered by CorpusIQ