New Hermes Agent Update IS INSANE!

New Hermes Agent update is insane. The new Hermes Agent VOR.20.0 update just dropped and the multi-agent kanban inside it is the wildest piece. You now have a board. Agents pull tasks off the board on their own.

They work side by side. They hand off when stuck. You watch the whole thing from one screen. No more sitting in 3 terminals trying to keep up.

Let me show you what this means in plain English. Before this, if you wanted Hermes to do 5 things, it did them one at a time. Or you opened 5 terminals, ran 5 copies and tried to track them yourself. Painful.

Slow. You'd lose your place. You'd forget which one was doing what. Now there's a real board.

It looks like the Trello or Notion board you already use. There are columns. Triage. Ready.

Running. Blocked. You drop a task in. Pick who works it.

The board does the rest. And the agents are not pretending to work together. They actually do. Each task is a row in a small database on your machine.

Every comment, every handoff, every update is a row anyone can read. A researcher agent can drop notes. A writer agent picks them up later. A reviewer steps in after that.

None of them have to be online at the same time. None of them lose context. That part matters. I'll come back to it.

First, the basics. Hermes Agent is a free, open source AI agent that lives on your computer. It can do real work. Send messages.

Run research. Pull data. Manage your inbox. Join meetings.

It plugs into the AI model you want. Claude. GPT. Gemini.

Kimmy. GLM. Whatever. You pick.

The new version came out April 30th, 2026. The note says 1096 commits and 213 people worked on it. It's a big drop, not a small patch. The headline feature is the Kanban.

The board for many agents. Here's how it works. You type one line in your terminal. Something like Hermes Kanban create write blog post about email funnel.

Researcher. That makes a card. You assign it to a profile. Profile is just a named agent.

You might have one called Researcher. Called Writer. Called Reviewer. Each one has its own setup.

Its own tools. Its own memory. Its own model if you want. Then a thing called the Dispatcher checks the board.

It sees a ready task. It picks it up. Spawns the right agent. The agent reads the task, the comments, and gets to work.

When it's done, it writes a result. The card moves to done. If it gets stuck, it moves to blocked and pings you. You answer.

It picks back up. You can do this with one agent. With 10. They run in parallel.

Real parallel. Not fake. I'll switch back and forth. Parallel.

A full separate process is on your machine. Now think about your own work. Most of you are probably running one AI agent at a time. You ask it to do something.

You wait. You ask the next thing. You wait. Half your day is waiting.

This kills the waiting. Quick word here. Inside the AI Profit boardroom, we're already deep on Hermes Agent. We've got a step-by-step setup walkthrough for Vioton 2.0.

Four coaching calls every week where we go through the Kanban live and a 30-day playbook on how to run a multi-agent setup for your business. Researchers pulling leads. Writers turning them into emails. Reviewers checking the output.

All on one board. Members in there. A lot of them already running Hermes. Prompt library built around it.

And a member map so you can find people near you doing the same thing. Link in the description or go to AIProfitBoardroom.com. Back to the update. The thing that makes this Kanban different from the other agent boards out there is the comments.

Every card has a comment thread. Agents read it. Humans read it. You read it.

So when a writer picks up a task, it can see what the researcher already found. It doesn't start from zero. It builds on what came before. This is huge.

Most AI agents have memory problems. They forget. Or they pile up old stuff. With the Kanban, the memory is on the card.

Right there. Easy to read. Easy to check. Easy to fix if something looks off.

There's also a thing called workspaces. Each card gets a folder. A scratch space. The agent works in that folder.

Doesn't touch your other files. Doesn't make a mess of your computer. When the task is done, the folder gets cleaned up if you want. Or kept if you want a record.

You pick. And every task can have a parent. A child. A whole tree of work.

So you can say research this topic and have that one task spawn five smaller tasks. Three researchers run at the same time. An analyst pulls their notes together. A writer turns it into a final piece.

The dispatcher only moves a task to ready when all the parents are done. So the writer doesn't start until the research is in. No mess. No double work.

You can also tag tasks by what they call a tenant. That's just a fancy word for client or project. So if you serve three businesses, you tag each task with the business name. The same agents handle all of them.

But the data stays separate. One translator handles content for client A and client B and never mixes them up. Let me give you one simple example you can use today. You run a small business.

You get five customer questions in your inbox every morning. You drop each one on the Kanban as a task. A support agent profile picks them up one by one, drafts a reply and puts it back on the card for you to check. You walk in with coffee.

Five drafts ready. You read, tweak, send. That's an hour of work done before you sat down. That's it.

That's the whole pitch. Less waiting, more done. Now let's talk about how this is different from what came before. Before this, Hermes had a thing called delegate task.

You could ask a sub-agent to do a quick job inside a chat. Useful, but it died at the end of the chat. No memory. No record.

Just a quick helper. The Kanban is the opposite. It survives. You close your laptop.

You come back. The board is still there. The work is still there. The history is still there.

Tasks that were running keep running. Tasks that finished are still marked done. Tasks waiting on you still wait. This is what they mean by durable.

The board doesn't care if you log off. It doesn't care if Hermes restarts. It doesn't care if your computer crashes. The data sits in a small file on your disk.

When everything comes back online, the work picks up where it left off. That alone changes how you can use this. You can leave a board running for weeks. Agents check in every day.

Build up a journal. Keep notes. Track progress. Hand things back to you when they need help.

Quick aside. There was a small bug with the first release. Board was tied to your active profile, so workers spawned by the dispatcher couldn't see tasks made from the main profile. Reported by users on April 30th.

The team is patching it. If you hit it, just keep one profile for now or grab the latest patch. Worth knowing. Okay, why does this matter for your business?

Most people running AI right now use it like a single tool. One chat. One question. One answer.

Like a fancy search engine. That's fine, but it's nowhere near what's possible. Multi-agent flips it. You stop being the worker.

You become the manager. You set the goal. You hand the goal to a board. The board figures out how to break it down.

Workers do the parts. You check the result. Done. Think about how a small team works.

You have a person who does research. A person who writes. A person who edits. A person who handles email.

They talk to each other through slack or a shared dock. They don't sit in the same room. They don't all work at the same time. But the work flows.

That's what the Hermes Kanban gives you. Except the team is AI. And the team works while you sleep. There's a real comparison out there.

The same kind of board exists in tools like Agent Kanban, OpenClaw setups, Paperclip, Nanoclaw. Hermes does it differently in one important way. Every worker is a full process on your computer with its own identity. Not a thread inside another agent.

Not a child of a parent that can crash and take everyone down. Each one stands on its own. If one fails, the others keep going. If one crashes, the dispatcher picks up the task again on the next round.

That's the kind of thing that doesn't matter on day one. But on day 30, when you've got a real workflow running every day, that stability is what keeps it alive. Now a fair word. This is not point and click yet.

You need to be okay running terminal commands. You need to set up profiles. You need to install the gateway. The docks are good.

The community is active. But it's not a polished app you download from the app store. It's a tool for people who want power and don't mind a little setup. That's where most people get stuck.

The setup. The where do I even start feeling. You stare at the terminal and freeze. One more thing to know about virtuoso.

The Kanban is the headline, but it's not alone. The same release added an autonomous curator. That's a background agent that cleans up your skill library every seven days. Old skills get pruned.

Duplicates get merged. You never have to think about it. Your agent gets smarter over time without you doing a thing. Startup time also got cut by more than half.

The terminal interface used to take a while to load. Now it pops up almost right away. Small thing, but if you use this every day, that's a lot of time back. There's also a new Google Meet integration.

Hermes can join your meeting, turn on captions, take a transcript, and send you a summary after. You don't have to be in the call. The agent handles it. You walk in later, read the notes, and act on what was decided.

Add it all up. The Kanban for many agents working together. The curator keeping things clean. Faster startup.

New meeting tools. New messaging support. And it's free. And it's open.

And the people building it ship every week. Six months ago you needed to babysit your AI agent. Watch it. Clean up after it.

Tell it the same thing twice. Manage its memory. Now the agent manages itself. The Kanban manages the work.

Multiple agents manage each other. You step back and check in. That's a real shift, not a small one. If you want help building your own Hermes Kanban setup with multiple agents working together, the AI Profit Boardroom has full walkthroughs, four weekly coaching calls where we troubleshoot your setup live, daily tutorials on Hermes and other AI agent tools, a 30-day playbook for the multi-agent board, prompt library, and a community of 2,800 members already running this stuff for their business.

Many of them on Hermes right now sharing what works. Member map included so you can find people near you doing the same. Link in the description or go to AIProfitBoardroom.com. And if you want the full process, SOPs, and 100 plus AI use cases like this one for free, join the AI Success Lab.

Links in the comments and description. You'll get all the video notes from there, plus access to our community of 68,000 members crushing it with AI. See you in the next one.