Hermes Atlas
Multi-agent & orchestration

Hermes-AgentMesh

seleman66eeddwegger3-art/hermes-agentmesh

Redis-backed asynchronous message bus for long-running tasks between Hermes agents on different machines

In short

Hermes-AgentMesh is a peer-to-peer, Redis-backed asynchronous message bus for passing long tasks between agents on different machines. It removes the synchronous HTTP waits that time out when Hermes Agent tasks run for many minutes.

What Hermes-AgentMesh does

Hermes-AgentMesh is a peer-to-peer asynchronous message bus built on Redis queues, written by a Hermes agent persona called Bobo according to its README. Instead of waiting on a synchronous HTTP call that times out after 5 to 10 minutes, a sender pushes a task to a node's inbox list and returns immediately, and a resident worker pulls it, calls the LLM over HTTP and posts the result to an outbox.

The design uses no cross-machine SSH. A central Mac mini hosts Redis, the Hermes API on port 8642 and a file server on port 8080, while other Mac, Linux, Raspberry Pi or WSL nodes run worker_node.py as systemd user services or LaunchAgents. An orchestrator script drives a four-step protocol for multi-round debates between agents. The README says it works with other frameworks such as OpenClaw, LangGraph, AutoGen and CrewAI, and documents a case that cut a 300 second timeout to 8 seconds. The README is written in Chinese.

Key features

  • Redis inbox and outbox queues per node instead of synchronous HTTP calls
  • Resident workers run as systemd user services or macOS LaunchAgents
  • No SSH needed between machines
  • Orchestrator script for multi-round agent debates
  • Reports land on the machine that started the orchestrator
  • Adapters described for OpenClaw, LangGraph, AutoGen and CrewAI

When to use it

  • Sending a task that takes 10 minutes or more to a Hermes agent on another machine
  • Running multi-round debates between agents on separate devices
  • Letting a worker survive a restart without losing queued tasks

Who it is for: People running several Hermes agents across home or lab machines who need reliable long-task hand-offs.

How it fits with Hermes Agent

It is built around Hermes Agent's API server on port 8642 and framed as infrastructure for Hermes agents, while remaining usable by other frameworks.

Requirements: A Mac mini running Redis and the Hermes Gateway on :8642, plus one or more Mac, Linux, Pi or WSL nodes on the same local network

Note: The quick start binds Redis to 0.0.0.0 with protected mode off and assumes all machines share one local network.

FAQ

What is Hermes-AgentMesh?

Hermes-AgentMesh is a Redis-backed asynchronous message bus for agents. Tasks are pushed into a node's inbox queue and picked up by a resident worker, so long jobs do not hit HTTP timeouts.

Does Hermes-AgentMesh work with Hermes Agent?

Yes. The central node runs the Hermes API on port 8642 and workers call it over HTTP. The README also says other frameworks can use the same bus.

Is Hermes-AgentMesh free and open source?

GitHub could not identify a standard license for this repository (it reports NOASSERTION), so check the repository and contact the author before reusing the code.

Similar multi-agent for Hermes Agent

All multi-agent

Related guides: Build agent teams with the Hermes kanban board · Run multiple Hermes agents with profiles · Connect Hermes agents on several machines with Hermes Desktop