Get Hermes Agent Up and Running in 10 Minutes: Step-by-Step Installation and Usage Guide to Build Your Own Ever-Smarter AI Assistant.
Video Tutorial: https://www.bilibili.com/video/BV1yQR8BhEhm/
Preface
Ever had that feeling where every time you open an AI chat window, you have to explain your identity, preferences, and work background all over again? Where the agent seems to “start from scratch” every time it takes on a task, forgetting all previous experiences so you end up fixing the same bug manually again and again?
What if you had an AI assistant that could deploy in just a few minutes, could remember every conversation you’ve had, turn those experiences into reusable Skills, and stay available 24/7 in WeChat, Feishu, or Telegram, always ready to respond—not just for a one-off task, but to continually learn your work style?
That’s what Hermes Agent offers.
As an open-source, self-hosted AI Agent system, Hermes supports major models like Kimi, GLM, Claude, Gemini, and more, but stands out for its “long-term memory + skill evolution” core design: it learns lessons, builds up skills, compresses context, and makes repeated tasks faster and more token-efficient. Compared with similar products, it’s more transparent in tool invocation, offers lower migration cost (you can switch from OpenClaw with just one command), and is truly designed for long-term task management.
In this guide, you’ll be taken step by step through environment preparation, model configuration, TUI chat setup, and message gateway integration to complete your first Hermes Agent deployment, hands-on.
Join the webmaster group: 767557452
Hermes Agent Overview
Hermes Agent is an open-source, self-hostable, lightweight AI Agent with long-term memory and skill-building abilities, optimized for Chinese users’ environments and needs. Core strengths include:
Ultra-flexible deployment: Supports local PCs, VPS, Docker, WSL2, and works across Linux/macOS/Windows, with mirrors for faster installation in China.
Always evolving: Remembers your projects, preferences, and work habits across sessions, turning solved problems into reusable Skills that make it increasingly useful.
Comprehensive tools: Integrates 40+ tools including MCP, terminal, file, browser, image, TTS, and can run automated reporting, backups, and inspections via cron.
Always online across platforms: Integrates with QQ, WeChat, Feishu, DingTalk, Telegram, Discord, etc. for anytime-anywhere response.
Fully model-compatible: Works with Qwen, GLM, Kimi, MiniMax, Claude, Gemini, OpenAI-compatible APIs, and local models, all optimized for Chinese networks.
HermesAgent official site: https://url.zeruns.com/HermesAgent
Preparation
For demonstration, I’ll use a Linux cloud server, but you can also deploy on your own Raspberry Pi, Mac, small PC, etc. A cloud server is often easier—if you want to host a website, for example, you can have your AI agent build and deploy it right away for public access.
Recommended cloud servers:
- Yuyun Hong Kong Five-zone high-bandwidth VPS review, 2C/2G 500Mbps, only 19.6 CNY/month: https://blog.zeruns.com/archives/932.html
- Cool Duck Data Hong Kong VPS performance test, direct three-network connection, 2C/2G 5M, only 6.9 CNY/month: https://blog.vpszj.cn/archives/2646.html
- Beishao Cloud Hong Kong A Zone VPS, 2C/2G 10Mbps, only 18 CNY/month, direct three-network connection: https://blog.zeruns.com/archives/933.html
- Other cost-effective servers recommended: https://blog.vpszj.cn/archives/41.html
I’ll use a Yuyun cloud server for this walkthrough. First, use the promo link or code (‘zeruns’) to register for a Yuyun account. Log in and go to Console, click Cloud Server → Purchase Cloud Server.
- Yuyun promo registration (AFF): https://www.rainyun.com/zeruns_?s=nodeloc
- Yuyun promo code: zeruns
- Registering with the promo code gets you a 50% off coupon for the first month. You can also claim a special 20% off coupon in the Points Mall, which can be combined with the official annual 30% off discount for a total of 44% off.
Then, choose your server region and configuration as needed. I picked Hong Kong Zone 2 with 2C/2G. If you want to host a website, Hong Kong/Japan/US is recommended for no ICP license hassle.
Pick Debian13 as the system, then click Buy Now or Try—just 1 yuan for a 1-day trial.
After purchase, your server will show up under My Cloud Servers. Click Manage.
You’ll now see the server info page, where you can reinstall/switch system or upgrade specs. Wait for server creation to finish before proceeding.
While the server is being set up, you can prepare the API for your LLM platform. Here are some recommended platforms:
- Ucloud AstraFlow (AFF): https://astraflow.ucloud.cn/modelverse/playground?cps_code=2DHPjTLpGvIG49HCNDWD0h
- Alibaba Cloud (get a 10% off coupon via this link) (AFF): https://www.aliyun.com/minisite/goods?userCode=jdjc69nf
- Silicon Flow (register with this link for a ¥16 voucher) (AFF): https://cloud.siliconflow.cn/i/hSviAP2x
I’ll demonstrate with Ucloud’s AstraFlow Modelverse. Register via the link below, then enter Ucloud Modelverse, and go to Key Management in the lower left, then Create API Key.
- Ucloud promo registration link (AFF): https://passport.ucloud.cn?cps_code=2DHPjTLpGvIG49HCNDWD0h
- Register via the link for 500,000 free tokens
Set any name for the API Key, optionally a budget, then click Confirm.
Copy and save the API Key—you’ll need it for the Hermes installation. Other platforms have a similar process.
Connecting to the Server
Download and open an SSH client; Putty or MobaXterm is recommended.
SSH client download: https://www.123pan.com/ps/2Y9Djv-UAtvH.html
I’m using MobaXterm. In the SSH client, enter your server’s IP address (from the console), SSH port (default is 22), then click OK or Open.
Enter the username and hit Enter, usually root. Then enter the password (from the console) and press Enter. The password won’t show as you type.
Tip: In the SSH terminal, select text with the left mouse button and release, then click somewhere blank to copy. Right-click to paste.
APT Source Mirroring (Skip for overseas servers)
By default, the system APT download sources are foreign servers, so you should switch to a mirror in China using chsrc.
In the SSH terminal, enter the following commands (lines beginning with # are comments; don’t input those):
# Download and install chsrc
curl https://chsrc.run/posix | bash
# Auto test speed and set fastest source for Debian
chsrc set debian
Installing and Configuring Hermes Agent
In the SSH terminal, enter the following command and press Enter:
curl -fsSL https://res1.hermesagent.org.cn/install.sh | bash
Wait for the installation to finish.
When you see the prompt as shown below, installation is complete and you’ll enter the configuration wizard. Press Ctrl+C to exit the setup wizard for now, then enter hermes setup in the terminal to start again.
Now you’ll be asked to choose between Quick Setup or Full Setup; just hit Enter to select the default Quick Setup.
Next, set the LLM API provider. Use the up/down (↑↓) arrow keys to select Custom endpoint (enter URL manually). Press Enter to confirm.
Now input the Ucloud API URL. In the Modelverse’s “Model Square”, choose any model and click Reference API to see the API docs and get the URL. The Chinese mainland API URL is usually https://api.modelverse.cn/v1. Enter, then hit Enter to confirm.
Then, enter your API Key (created earlier). It won’t show as you type. Hit Enter to confirm.
Now choose an AI model. Usually the system will fetch the list from the API; just enter the number or ID, or copy the ID from the Model Square. For example, deepseek-v4-flash is recommended for AI Agent scenarios—good cache hit rate and very cost-effective.
Next, set context length—just hit Enter for auto-detect.
Set a display name—hitting Enter accepts the default.
Now message platform setup, press Enter to set up now.
Select a message platform and use the spacebar to select—I chose QQ Bot, then press Enter.
Select the first item for QR code automatic setup, and press Enter.
Copy the link provided, open it in your browser, and a QR code will appear. Scan it using the QQ app and follow the guide to create a QQ bot—you’ll communicate with Hermes via this bot.
Set up message authorization; just keep pressing Enter for defaults.
Now set the message gateway to run in the background; select the second option to run as a system service, then Enter.
Set which user to run the gateway as—input root and press Enter.
Press Enter again to start the service.
Press Enter one more time to enter terminal chat—you can also use the QQ bot to talk to Hermes.
Hermes Agent is now fully deployed and configured! You can use it immediately. You can even send new multimodal API links via QQ, letting Hermes configure them on its own for functions like image recognition.
Using Hermes Agent
Next, I’ll send commands to Hermes on the server via the QQ bot, instructing it to design and deploy a Minecraft server homepage.
Here’s what the site it made looks like—pretty impressive! You can keep sending instructions for further improvements.
For features like file sending/receiving via QQ bot, you can have Hermes update its code, enabling you to send files directly or have it send files back. You can also have it add the DeepSeek-V4-Pro model for complex task planning.
Some system prompts are in English—you can get Hermes to localize itself. There’s lots of fun to be had!
That’s the end of this tutorial, but there’s plenty more to explore!
Recommended Reading
- Great value VPS/cloud server recommendations: https://blog.zeruns.com/archives/383.html
- Minecraft server hosting guide: https://blog.zeruns.com/tag/mc/
- Discourse forum setup guide, zero to self-hosted open-source community: https://blog.zeruns.com/archives/919.html
- 5G CPE Kunpeng C2000MAX quick unboxing review: https://blog.zeruns.com/archives/937.html
- CommentAI-Typecho blog AI comment moderation/reply plugin: https://blog.zeruns.com/archives/931.html
- Op-Amp (operational amplifier) beginner’s guide, from basics to hands-on: https://blog.zeruns.com/archives/938.html
English Version of the Article: https://blog.zeruns.top/archives/90.html