Hermes Agent Swarm: Free Multi-Agent Update

The Hermes agent swarm feature is the free Hermes Workspace update that turns one AI agent into an entire team. After running it for a few weeks I'm convinced this is the upgrade that takes Hermes from "useful single agent" to "real production multi-agent system" — and it's all free.

Single-agent workflows are slow because each step happens one at a time and long tasks take ages. Hermes agent swarms fix that by running multiple agents in parallel — one plans, one builds, one reviews, all at once and all coordinated automatically. This post is the full walkthrough.

What Hermes Agent Swarm Is

A new feature inside Hermes Workspace that's free and was created by community contributor Eric. It lets you run a team of Hermes agents simultaneously, where each agent has a role, a model, a mission, and memory.

The orchestrator (called Aurora) routes work between the agents. You give the team a goal and it figures out how to get there without you needing to micro-manage the steps.

What's New In This Update

Three things to know about the latest update.

The first is agent swarms themselves, which let you run multiple agents in parallel. The second is phone access, so you can control your swarm from mobile. The third is auto-routing — give Aurora a goal and she composes the routing plan herself.

This is one of the biggest free updates Hermes Workspace has shipped.

Setting Up Hermes Agent Swarms

Five steps to get a working swarm running.

Step one is installing the plugin — use the free plugin Eric created and run the install command in your terminal to set up Hermes Workspace. Step two is running the Hermes gateway, which is the standard Hermes startup. Step three is starting the workspace UI by connecting to Hermes Workspace on your local host. Step four is updating if you already had Hermes Workspace — click Update first, because the swarms feature only works on the latest version. Step five is opening the Swarms section from the left sidebar in Hermes Workspace, where you'll see workers for planning, routing, reports, and more.

🔥 Want my full Hermes Agent Swarm setup? Inside the AI Profit Boardroom, I share my Hermes Workspace configs, swarm prompts, and the 30-day road map for using swarms in real workflows. Plus a 2-hour Hermes course and weekly live coaching. 3,000+ members. → Get the setup

Adding A Swarm Agent

In the Swarms section, click "Add swarm" at the top right and pick a role from the presets.

The available roles are Builder, Reviewer, Triage, Lab, Sage, and Scribeex. Each role has preset system prompts and skills tuned for the job. If none fit, click "Custom" for full control over the system prompt, mission, and skills assigned to that agent.

The Main Agent (Aurora)

At the top of your swarm sits Aurora, the orchestrator. Her job is to receive your mission, decide which sub-agents handle what, route work between them, and compile results.

You give Aurora a mission. She compiles a routing plan. The team executes. That's the whole loop, and it works surprisingly well in practice.

Real Mission Example

From the demo, a real mission would look like: "Create a new AI automation blog for SEO + plan keyword strategy + create the content. I want to rank for AI agent community related terms + send traffic to [URL]."

Aurora composed a plan where the research agent gathered keyword data, the strategy agent planned the content calendar, builder agents created the actual blog posts, and reviewer agents checked the work. The result was a full blog with keyword research, competitor analysis, content brief, templates, and a 90-day content plan — built in 5-10 minutes.

That's the power of the swarm.

What You See In The UI

The Swarms section gives you a rich UI that's a real upgrade over managing individual terminals.

You get an Office view that shows every agent in either round table or grid layout. Per-agent terminals let you see what each agent is doing in real time. File outputs are saved locally for every piece of generated content. A browser preview opens generated work directly in your browser. Blocked vs ready agents are instantly visible so you can see what needs attention.

Compare this to managing 5 separate Hermes terminals and you'll understand why the upgrade matters.

Phone Access

Hermes Workspace's phone access works for swarms too. You can trigger a swarm from your phone, check progress on the go, and approve outputs remotely.

Your agent team is running 24/7 in your pocket. This pairs naturally with OpenClaw Aion UI's phone access if you're running both.

When To Use Swarms (And When Not)

Use swarms when the task has multiple distinct phases (research, write, review), output quality matters, you want speed via parallelism, or the mission is too big for one agent.

Skip swarms when the task is small, you're prototyping, or you're doing single-step work. For one-off chats, single-agent is fine. For production work, swarms are the upgrade.

Common Setup Issues

Three issues that trip people up most often.

The first is an agent showing "blocked." Click the agent and check its settings — usually it's the wrong API selected, so switch to a model the agent can actually use. The second is plugin install failing. If you're not technical, ask Claude Code to install for you — this trick works for most Hermes installs and is covered in Free Claude Code. The third is the swarms feature being a bit buggy. If you hit issues, ask Claude Code to debug — it communicates with Hermes well and usually finds the problem fast.

Models For Each Swarm Role

Don't put every agent on the same model — match the model to the role.

For the researcher, use a long-context cloud model like Kim K2.5. For the builder, use a creative cloud model that produces good first drafts. For the reviewer, use a small fast model that's cheap to run on every piece of output. For sub-agents in larger swarms, Nvidia Nemotron 3 Nano Omni is designed specifically for that role.

I cover model picks in Hermes Gemma 4 and Hermes DeepSeek.

Daily Reality

What it looks like running swarms daily.

In the morning, I give Aurora a content mission. Aurora spawns the swarm and multiple agents work in parallel. 10-15 minutes later, the full output is ready and I review and publish.

What used to be a half-day task is now a coffee.

Why Swarms Beat Single-Agent

Three reasons swarms produce better outcomes than running one agent.

The first is parallelism — multiple tasks happen simultaneously, so you're not waiting for one step before the next starts. The second is specialisation, where each agent does one thing well rather than a generalist trying to do everything at average quality. The third is scale, because adding more agents scales output without proportional time investment.

🚀 Want my full Hermes swarm playbook? The AI Profit Boardroom has my Hermes swarm configs, 2-hour Hermes course, daily training drops, and weekly live coaching. 3,000+ members building real automations. → Join here

FAQ — Hermes Agent Swarm

Is Hermes Agent Swarm free?

Yes — free plugin for Hermes Workspace.

Do I need Hermes Workspace installed first?

Yes — swarms run inside Workspace.

How many agents can I run in a swarm?

Hardware-dependent — I run 4-6 agents comfortably.

Can I run swarms from my phone?

Yes — Hermes Workspace supports phone access.

Will swarms work with cloud models?

Yes — but local Ollama models are cheaper for high-volume swarms.

What if a swarm agent gets stuck?

Check its API settings. Most stuck agents are model misconfiguration.

Can I create custom swarm roles?

Yes — fully customisable system prompts, missions, and skills.