Hermes AI Agent Setup
2026-11-10
Your own AI agent, set up and working. 21Labs installs, configures and secures a Hermes AI agent in your own environment, connects it to the tools your business already uses and gets your team using it. We are an agency in Sydney that serves businesses across Australia.
This page names a specific product. Elsewhere on the site we say "AI tools". Here the product is the subject, so we name it.
What Hermes Agent is
Hermes Agent is an open-source AI agent built by Nous Research and released under the MIT licence. Nous Research describes it as a self-improving agent. An agent differs from a chat window in that it can work through a task with tools, such as running commands, searching the web and reading files, instead of only replying to a message.
These are the features the project documents:
- Memory and skills. It keeps memory across sessions, can search past conversations, and can turn what it learns from a hard task into a reusable skill that it refines with use.
- Messaging. One gateway process lets you reach the agent from the command line and from messaging platforms such as Telegram, Discord, Slack, WhatsApp, Signal and email.
- Scheduled jobs. A built-in scheduler runs tasks on a timetable and delivers the results to a platform you choose.
- Subagents. It can start separate subagents to work on parts of a job in parallel.
- Tool connections. It can connect to MCP servers, a common way of giving an agent access to other software.
- Your choice of model. It works with a range of AI model providers and custom endpoints, and you can switch between them.
- Places to run. The documented terminal backends are local, Docker, SSH, Daytona, Modal and Vercel Sandbox, and it installs on Linux, macOS, Windows and Android.
Everything above comes from the project's own repository and documentation, linked at the end of this page. The project moves quickly, so we check the current documentation at the start of each setup.
What running it in your own environment involves
Because Hermes Agent is software you run, someone has to make real decisions before it is useful.
Where it runs. On a laptop it stops when the laptop sleeps. On a small server or a cloud account it stays on, and your team can reach it through the messaging apps they already use. Each choice changes cost, privacy and who looks after it.
Which AI model it uses. The agent needs an AI model to think with. The model provider sees what the agent sends it, and the provider's charges are separate from the software. This is a business decision about cost and data.
What it can reach. An agent that can run commands and read files can also do damage. The project documents layers of protection: approval before risky commands, rules for who may talk to the agent, protected file paths and isolated containers. The documentation also says these reduce accidental harm and do not sandbox a hostile agent on a local setup. A good setup therefore limits access to what each job needs.
How the team will use it. Software that nobody opens delivers nothing. Someone has to show people which jobs to hand it and where to draw the line.
Who it suits
Hermes setup suits you if you recognise some of these:
- You want an AI agent working in your business and want to control where it runs.
- You are comfortable with the idea of an agent but not with installing and securing one yourself.
- You have tried to set one up and stalled on configuration or connections.
- You want the agent reachable from the messaging tools your team already uses.
- You want someone to hand it over in a state your team can use on day one.
If you can say what you would like an agent to help with, you are ready for discovery.
What 21Labs does
Choose where it runs
We decide with you where the agent lives, from hardware you own to a cloud account you control. We explain the trade-offs and record the decision.
Install and configure it
We install Hermes Agent, choose the model provider with you and configure the agent for your business. We document the configuration so it can be reviewed, repeated and changed later.
Connect it to your tools
We connect the agent to the places your work happens, such as your team's messaging apps and the software your business runs. We give it the access the jobs need and no more.
Secure it
We turn on the available protections, limit what the agent can reach, handle credentials with care and write down rules for your team on what to share with it.
Hand over and support
We walk your team through using the agent on real jobs from your business. On a monthly retainer, 21Labs supports the setup as your tools and needs change.
How the engagement runs
Every 21Labs engagement runs the same cycle.
1. Discovery is free
Start with a free discovery. We learn your business and come back with a plan and a price. For a Hermes setup, discovery settles what the agent should help with, where it runs and what it connects to.
2. Price agreed before work starts
You agree the price before we start, so there are no surprises. Instalments are allowed and the total stays fixed.
3. Delivery
We install, configure, connect and secure the agent, then hand it to your team with a walkthrough.
4. Measured results
We agree measures of success in discovery, such as how many people use the agent each week or how long a chosen task takes with it. After delivery we measure them and show you what changed.
5. Renewal on your terms
After delivery, we stay with you each month to keep improving the results, or you stop there. Timing is agreed with you in discovery, because it depends on where the agent runs and what it connects to.
Where to start
Discovery is free, and it is the best first step. If you are not sure which size fits, get in touch and we will work it out together. The packages and prices are on the services page.
Who it is not for
- A business that wants a hosted service with nothing to run. This setup puts the agent in your own environment, which needs a decision about where.
- A team with a manual process to remove. If a fixed set of steps should simply run on its own, workflow automation is the better fit.
- Someone who wants a result promised in advance. We set the agent up and measure how it is used. What it achieves depends on the jobs you give it.
- A business that has not decided what an agent should do. Start with a free AI readiness report or an AI strategy, which finds where AI is worth the effort.
Sources
- Nous Research, Hermes Agent repository, read 10 November 2026
- Nous Research, Hermes Agent documentation, read 10 November 2026
- Nous Research, Hermes Agent security page, read 10 November 2026
Start with a free discovery
Tell us what you would like an agent to help with. We will come back with a plan and a price for setting up Hermes Agent in your environment, and nothing starts until you have agreed it.
Frequently asked questions
- How much does a Hermes setup cost?
- Discovery is free. After it you get a scoped proposal with a price you agree before any work starts. Larger or more involved setups are quoted after discovery. The packages and prices are on our services page.
- How does it start?
- You send the contact form and tell us what you would like an agent to help with. We reply within 24 hours and book a free discovery. It ends with a plan and a price.
- Is Hermes Agent free?
- The software is open source under the MIT licence, so there is no licence fee. Running it still has costs, such as the machine or cloud account it runs on and the AI model provider you connect it to. We estimate both with you in discovery.
- Where does it run, and who controls the data?
- The Hermes documentation lists several places to run it, including your own machine, a server you reach over SSH and a Docker container. We choose with you in discovery. Wherever it runs, the agent sends your requests to the AI model provider you configure, so we also settle which provider and what the agent may share.
- Can it do anything dangerous on our systems?
- It can run commands, which is why security is part of the setup. Hermes has documented controls such as human approval for risky commands, allowlists for who may talk to it and container isolation. Its own documentation says these reduce accidents and do not make a hostile agent safe on a local setup, so we limit what it can reach.
- Do I need technical staff to run it?
- No. We handle the installation and configuration and hand over with a walkthrough. On a retainer we support it afterwards, so your team does not need to maintain it.
- Is Hermes the right agent for my business?
- We check in discovery. If another approach would serve you better, such as an automated workflow, we say so. We would rather match the tool to the job than sell you a setup you do not need.
