How to use and build Hermes Agent | Installation, model settings, and Gateway collaboration
Hermes Agent is an AI agent published by Nous Research. You can install it on a desktop or through a command line, choose an AI model, and start chatting. This article covers installation, model setup, operation checks, and optional messaging integration. The steps are based on the official Quickstart; they are not results of hands-on testing.
official Quickstarthttps://hermes-agent.nousresearch.com/docs/getting-started/quickstart
How to use Hermes Agent will be easier to understand if you check the following steps: ``Installation, Model Selection, Initial Conversation, and Resuming Conversation.'' There is no need to set up Gateway integration such as Telegram or periodic execution from the beginning. You can start with Desktop for Mac and Windows, and CLI for Linux and terminal-based users.
If you want to build Hermes Agent for business use, including model settings and cooperation with external services, please see ZETTAi's AI agent construction support.
ZETTAi's AI agent construction supporthttps://zettai-ai-agent-site-render.onrender.com/
How to use Hermes Agent: Quick guide to the first conversation
| What to do | When using Desktop | When using CLI |
|---|---|---|
| Introduction | Insert Desktop package from official website https://hermes-agent.nousresearch.com/docs/getting-started/installation | Execute official installation procedure https://hermes-agent.nousresearch.com/docs/getting-started/installation that matches the OS |
| Model selection | Select the provider to use during initial setup | Select using hermes setup or hermes model |
| First conversation | Send one request that is easy to check in the app | Open hermes or hermes --tui and send one |
| Check the results | Check the model, reply, and allowed operations | After replying, check if you can resume the conversation with hermes --continue |
This table organizes the steps of Official Quickstart. The display and setting locations differ between Desktop and CLI, so proceed according to the method you use. If you have already installed the CLI, you can select a model and start a conversation with the following command:
Official Quickstarthttps://hermes-agent.nousresearch.com/docs/getting-started/quickstart
hermes model
hermes --tui
The first request should be something that allows you to check the results yourself, such as ``List the file names in this verification folder.Do not change or send them externally, and also tell me what files you cannot read.'' Once you get a response, exit and check if you want to continue from where you left off with hermes --continue. If a model or tool does not work properly, check with hermes doctor before adding Gateway.
What are “memory” and “skills” useful for?
If you can keep the premises and procedures that you use repeatedly, you can reduce the burden of rewriting long instructions each time. Skills are a mechanism for reusing how to proceed with specific tasks. However, what is saved is not always correct. It is necessary to review the operation when the person in charge or company rules change.
For example, if you are organizing articles that have been published internally, you can decide on the reference destination, classification criteria, and output format and try it out. Operations such as sending external emails or deleting files are separated into separate approval stages.
Choose how to install Hermes Agent
| Usage environment | How to choose first |
|---|---|
| Windows | Install Official Desktop package https://hermes-agent.nousresearch.com/docs/getting-started/installation. Can also be installed using only the command line |
| Mac with Apple Silicon | Official Desktop package or CLI introduction |
| Linux/WSL2 | CLI steps in the official installation guide |
The official Desktop package includes the agent execution environment. The macOS Desktop installer is for Apple Silicon. Since the distribution format and update conditions are different for Intel Mac, please check the official supported environment.
official supported environmenthttps://hermes-agent.nousresearch.com/docs/getting-started/platform-support
How to create Hermes Agent: Install with CLI
Below are the steps for CLI posted in Official Installation. Please check the latest command and compatible environment on the official page before execution. The installer retrieves the necessary files and begins an interactive initial configuration.
Official Installationhttps://hermes-agent.nousresearch.com/docs/getting-started/installation
For macOS/Linux/WSL2
Run the following in your terminal: The official guide instructs you to prepare Git, curl, tar, and SHA-256 tools for installing a POSIX environment.
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
For Windows
Run the following in PowerShell: If you want to use the desktop app, you can choose the Windows package from the official website.
official websitehttps://hermes-agent.nousresearch.com/
iex (irm https://hermes-agent.nousresearch.com/install.ps1)
If the command is not found after installation, reopen the terminal and check the PATH settings in the official guide. If you have already completed the initial setup, follow the steps below to check the model and conversations.
How to use Hermes Agent: up to the first conversation
- Install using the method appropriate for your OS from the table above.
- Run
hermes setupon the CLI and follow the instructions to set up the AI model provider. If you choose the connection destination yourself, you can use thehermes model. Also check the model terms and conditions and costs here. If you choose a local model, also check the required context length in the official documentation. - Start chatting with
hermesorhermes --tuiif you are using a terminal with a screen. - Make a request that makes it easy to check the results, such as ``Please list the file names in this verification folder and tell me the main points without changing them.'' Verify that the selected model is displayed and responds without error.
- Before expanding your work, make sure that you can resume the conversation with
hermes --continueand that only the tools you allow are used.
The command to select a model and start a conversation in the CLI is: If you did not complete the settings during installation, run hermes setup first.
hermes model
hermes
If the selected model is displayed on the screen and you answer the simple questions without errors, the basic operation check is complete. Check the contract terms and conditions, as usage fees may apply depending on the model and connection destination. After ending a conversation, you can resume the previous conversation with hermes --continue.
If you want to use the message app all the time, proceed with hermes gateway setup after the basic chat is working. If you set Gateway, periodic execution, and skill at the same time from the beginning, it will be difficult to isolate the cause of the problem.
Steps to build Hermes Agent: Always ready for use
"I was able to install it" and "I can use it for work" are two different states. Check the following four steps according to the order of Official Quickstart. Here, we assume a verification environment in which one person uses the CLI, and distinguish it from the results of running it on an actual device.
Official Quickstarthttps://hermes-agent.nousresearch.com/docs/getting-started/quickstart
| Stage | Operation | What determines completion |
|---|---|---|
| Model connection | Select a provider with hermes setup or hermes model | One round trip conversation is established with hermes |
| Scope of work | Check the tools used in hermes tools | Make sure that unauthorized operations do not occur due to read-only requests |
| Message coordination | Execute hermes gateway setup only when necessary | Can send and receive test messages from connected channels |
| Operation confirmation | Check the settings and operation with hermes doctor and hermes gateway status | Find out how to connect after restart, how to stop, and who is in charge |
Before connecting to an external messaging app, decide who will store your bot's token and which users will accept messages from it. If the server is to run constantly, the scope of construction also includes authentication information, logs, backups, updates, and restart procedures in the event of a failure. Please refer to Official Quickstart Sandbox Description for options to separate the execution environment using Docker or SSH.
Official Quickstart Sandbox Descriptionhttps://hermes-agent.nousresearch.com/docs/getting-started/quickstart
Example of first request and result confirmation
List the file names in this verification folder. Please do not change the content or send it externally, and also indicate the file that could not be read and the reason at the end.
After execution, a human verifies that the list matches the actual folders, that there are no unauthorized changes, and that the reason for failure is known. Rather than just looking at the deliverables and assuming that the process is complete, you can also record the models used, connection destinations, and allowed operations, so you can reproduce the settings later.
If it's installed but doesn't work
| Symptoms | Things to check |
|---|---|
hermes cannot be found | Reopen the terminal and check the PATH settings in Official Guide https://hermes-agent.nousresearch.com/docs/getting-started/installation |
| Model is not responding | Check provider, model name, and credentials with hermes model |
| Unable to resume previous conversation | Check session and profile in use with hermes sessions list |
| Messages are not delivered to Gateway | Check the status with hermes gateway status and review the token and connection destination settings |
If you cannot narrow down the cause, run hermes doctor and check the displayed settings. For local models, check the endpoint, model name, and context length according to Official Quickstart.
Official Quickstarthttps://hermes-agent.nousresearch.com/docs/getting-started/quickstart
Check items when using at work
Refer to the Official Security Guide to organize the commands that can be executed, directories that can be read, secret information that can be stored, and connections to external services. When AI reads a web page or document, it checks the operation results to ensure that the instructions written there are not treated as business instructions.
Official Security Guidehttps://hermes-agent.nousresearch.com/docs/user-guide/security
| Work suitable for initial verification | Work that requires careful design |
|---|---|
| Summarizing public information, organizing fictitious data | Modifying production files, sending to customers |
| Working with local exam folders | Payment, publishing, deletion, and changing authentication settings |
Differences between OpenClaw and cloud type
OpenClaw is a configuration that connects channels and tools centered on Gateway. Grok Bot and OpenAI dots have usage formats that use cloud computers. When choosing one, consider where it will run and who will manage permissions and logs, rather than the product name.
FAQ
Is Hermes Agent free? Public repository licenses and costs for models, servers, and operations are separate. Decide on the configuration and then make an estimate.
Can I store company secrets in the memory function? Do not enter until you have confirmed the storage location, access rights, and deletion method. First, try using fictitious information.
Can Hermes Agent be used on Windows? The official documentation provides instructions on how to install the Desktop package for Windows and native CLI. Please check the distribution format and compatibility conditions at Official Installation.
Official Installationhttps://hermes-agent.nousresearch.com/docs/getting-started/installation
Is Gateway required for Hermes Agent? no. You can try the first conversation on Desktop or CLI. When using from the Messages app, proceed with hermes gateway setup after the basic conversation has started.
If you would like to verify the AI agent at your company, please organize the target operations and necessary permissions and Contact us.
After trying the first conversation, if you want to proceed to the design of the environment and permissions that will be used regularly, you can check the support contents in Support for building and operating AI agents including Hermes Agent.
Support for building and operating AI agents including Hermes Agenthttps://zettai-ai-agent-site-render.onrender.com/