Keywords: Hermes Agent deployment, self-evolving AI agent, Tencent Cloud Lighthouse, cloud AI agent tutorial 2026, open-source AI agent
If you have been following the open-source AI community lately, there is one name you simply cannot miss: Hermes Agent.
It is not a luxury handbag, but the second most explosive open-source autonomous AI agent to appear in 2026, following the wildly popular OpenClaw ("Lobster"). In less than two months after its launch, Hermes Agent has surpassed 66,000 stars on GitHub and topped the global open-source rankings for several consecutive days. On April 14, 2026, Tencent Cloud's Lighthouse became the industry's first cloud platform to launch an exclusive application template for Hermes Agent, enabling one-click cloud deployment.
You might ask: "OpenClaw is already so powerful, so why should I pay attention to Hermes?" Because Hermes is not just another AI chat tool—it is a digital employee that genuinely self-evolves. The core concept is simple: you don't need to train it; it trains itself. From the moment you start using it, it quietly learns, remembers, and grows in the background.
More importantly, deploying a 24/7 cloud-based AI agent is far less complicated than you might think. This tutorial will walk you through the entire process from zero, using a Tencent Cloud Lighthouse server, so you can have your own "Hermès" up and running in about 15 minutes.
Who this tutorial is for:
- Developers and beginners who want their own dedicated AI agent
- Users who want to deploy Hermes Agent in the cloud for 24/7 availability
- Anyone looking to experience the hottest open-source AI framework at low cost
Estimated time: 10–15 minutes
Hermes Agent is an open-source autonomous AI agent developed by Nous Research, released under the MIT license. Its core is not simple chat conversations but "persistent memory + a complete self-learning loop"—it can autonomously extract skills from tasks and recall them across sessions, truly becoming smarter the more you use it.
In other words, the more tasks you give it, the more "experience" it accumulates. The next time it encounters a similar scenario, it no longer starts from scratch but directly invokes the capabilities it has refined before.
| Capability Module | Specific Functions | What It Can Do for You |
|---|---|---|
| Information Processing | Web search, content extraction, academic paper retrieval | Research assistance, information aggregation |
| File Operations | Read/write files, content search, code patching, format conversion | Document processing, code development, data cleaning |
| Terminal Execution | Run shell commands, manage background processes, control permissions | Server maintenance, automation scripts, task scheduling |
| Browser Automation | Web navigation, page interaction, visual analysis, form filling | Web scraping, automated testing, data collection |
| Multi-Platform Integration | Enterprise WeChat, Feishu, QQ, WeChat mini-programs, and 15+ other platforms | Real-time interaction, task delegation, result notification |
| Self-Evolution | Skill creation, self-optimization, memory retention | Long-term use, personalization, capability iteration |
These capabilities cover three major scenarios: personal assistance, development collaboration, and automated operations.
Many people ask, "Can't I just run it on my own computer?"
Frankly, local deployment has several serious drawbacks:
When deployed in the cloud, all these problems disappear. Cloud deployment achieves secure isolation from your local environment while providing 24/7 online availability. To put it simply: local running is a "toy," cloud running is a "tool."
Before you start, make sure you have the following:
| Item | Description | Required |
|---|---|---|
| Tencent Cloud account | Registered and identity-verified | ✅ Yes |
| LLM API Key | From providers like DeepSeek, MiniMax, OpenAI, etc. | ✅ Yes |
| Messaging platform account | Enterprise WeChat, QQ, Telegram, etc. | ⭐ Recommended |
| Payment method | Credit card, Alipay, WeChat Pay | ✅ Yes |
| ~10–15 minutes | The entire deployment process | ✅ Yes |
Note about API Keys: Hermes Agent does not come with its own LLM—it needs to connect to a third-party model API to work. For beginners, DeepSeek or MiniMax is recommended—easy registration and affordable pricing. If you just want to try it out first, you can also get a free model API key on OpenRouter.
This is the best deployment method for beginners—no coding required, just a few clicks.
Open your browser, go to the Tencent Cloud website, click "Log In" in the top right corner, and sign in with your registered account.
Don't have an account? Click the "Sign Up" button, follow the prompts to complete registration and identity verification—the whole process takes about 2 minutes.
After logging in, type "Lighthouse" in the top search bar, or find "Lighthouse" in the product list and click to enter. Then click "Create" or "Buy Now" to go to the server creation page.
This is the most critical step. On the creation page:
Why this image? Tencent Cloud Lighthouse is the industry's first cloud platform to offer an exclusive application template for Hermes Agent. Selecting this image automatically pre-installs all the environments Hermes Agent needs to run, saving you the trouble of manually pulling the source code and configuring dependencies.
| Configuration Item | Recommended Choice | Notes |
|---|---|---|
| Region | Choose the region closest to you | For global users, choose the nearest region (e.g., Singapore for Asia Pacific) |
| Package Specifications | 2 cores 4GB or higher | This is the official recommended configuration for Hermes Agent |
| System Disk | Default configuration is fine | Typically 60GB SSD is sufficient |
| Duration | Starting from 1 month | For first-time purchase, it is recommended to choose a 1-month trial |
| Server Name | Customizable, e.g., "my-hermes" | For easy identification |
| Login Credentials | Automatically generated password | Can be modified later |
Heads-up: It is not recommended to choose the 2-core 2GB entry-level package, as insufficient memory may cause the agent to run unstably. 2 cores 4GB is the official recommended minimum configuration.
After verifying that the configuration information is correct, click "Buy Now" → "Submit Order" → complete the payment. The server will be automatically created within 1–2 minutes.
About pricing: In 2026, Tencent Cloud Lighthouse offers significant new user discounts. It is recommended to check for free trials first before making a purchase.
After the server is created:
If you want to use an SSH client (such as Xshell, Termius) for remote connection later, you need to either "Reset Password" in the console to set a login password, or bind an SSH key under "Manage Keys."
If you already have a Lighthouse server (for example, one running WordPress or another application) and want to switch to Hermes Agent, it is also very simple.
Steps:
⚠️ Important Note: Reinstalling will erase all data on the server. Please be sure to back up any important content in advance!
If you have previously used OpenClaw, Hermes Agent includes a built-in
hermes claw migratecommand that can migrate OpenClaw settings, memories, skills, and API keys with a single command, making the transition very low-cost.
Once the server is up and running, the Hermes Agent runtime environment is ready, but the most critical part is still missing—the LLM capability. We need to configure a "brain" for it.
Log in to the server using the method described in the previous section, and you will see a command-line interface.
Enter the following command in the terminal to start the interactive configuration wizard:
hermes setup
The system will prompt you to choose a configuration method—it is recommended to select "Quick setup" and follow the prompts.
Hermes Agent is compatible with mainstream LLM APIs, including MiniMax, DeepSeek, OpenAI, Anthropic (Claude), OpenRouter, and others.
How should beginners choose? Two recommended options:
- Option A (Free trial): Choose OpenRouter → Register for a free account → Get an API key → Select a free model in the configuration (e.g.,
nvidia/nemotron-3-super-120b-a12b:free)- Option B (Stable): Choose DeepSeek → Register on deepseek.com → Get an API key → Follow the prompts to configure
After selecting the LLM provider, the system will prompt you to enter the API Key. Paste the API key you obtained from the corresponding platform (right-click to paste or Ctrl+V), and press Enter to confirm.
After the API key is verified, the system will list all the models supported by that provider. Select the model you want to use and press Enter to confirm.
⚠️ Important reminder: Hermes Agent recommends a context window of at least 64K to ensure sufficient working memory. Most mainstream models meet this requirement.
After configuration, the system will automatically test whether the model is working properly. You can also manually test it with the following command:
hermes
Then enter a message in the chat interface that appears to see if you receive a reply from the AI.
After configuring the model, Hermes Agent is already capable of working, but it can only chat in the terminal, which is not very convenient. Next, we will connect it to messaging platforms so you can interact with it anytime via WeChat, QQ, or Telegram.
Hermes Agent supports Telegram, Discord, Slack, Enterprise WeChat, QQ, Feishu, WeChat mini-programs, and 15+ other platforms through its unified messaging gateway.
Step 1: Register an Enterprise WeChat organization
If you do not have an Enterprise WeChat organization yet, go to the Enterprise WeChat official website and complete the registration.
Step 2: Create an Enterprise WeChat bot
Create a bot in the Enterprise WeChat management backend. After completion, obtain the bot's Bot ID and Secret.
Step 3: Write the credentials into the configuration
Return to the Lighthouse server terminal and edit the configuration file:
nano ~/.hermes/.env
Fill in the obtained Bot ID and Secret in the corresponding fields, save, and exit (Ctrl+X → Y → Enter).
Step 4: Start the gateway service
Execute the following command to register the gateway as a background resident service:
hermes gateway install
Once the gateway is started, your Hermes Agent is officially online! You can now send messages to your bot in Enterprise WeChat to test it.
Note: The configuration process for other platforms is similar—just select the corresponding platform in the
hermes setupconfiguration wizard and follow the prompts.
After deployment and configuration, here are some common commands you might find useful:
| Command | Description |
|---|---|
hermes |
Start interactive chat |
hermes model |
Switch LLM provider and model |
hermes tools |
Configure which tools to enable |
hermes config set <key> <value> |
Set a single configuration item |
hermes gateway |
Start the messaging gateway |
hermes setup |
Rerun the complete configuration wizard |
hermes claw migrate |
Migrate configuration from OpenClaw |
hermes update |
Update to the latest version |
hermes doctor |
Diagnose issues |
Enter hermes in the terminal and press Enter. Once you see the welcome screen, enter what you want the agent to do. For example:
Hermes Agent will call the appropriate tools (web search, file operations, code execution, etc.) to complete your request based on your instruction.
As you interact more with the agent, it will automatically extract reusable skills from tasks it has successfully completed. You can view them with the following command:
hermes skills list
This is the core mechanism behind Hermes Agent's "smarter the more you use it" feature—it continuously learns through practical use and builds up a dedicated skill library for you.
Q1: After server startup, the terminal says hermes: command not found.
A: The installation script may not have finished running, or the shell configuration may not have been refreshed. Try:
source ~/.bashrc
# or
source ~/.zshrc
Then run hermes --version to check if the installation was successful.
Q2: What if the model API call fails?
A: First, check whether the API key is correct and whether your account has sufficient balance. Then run the diagnostic command:
hermes doctor
This command will automatically check for configuration issues and provide suggestions for fixing them.
Q3: Can Windows users deploy?
A: Hermes Agent does not currently support native Windows environments. However, Windows users can install WSL2 (Windows Subsystem for Linux) first and then run it within WSL2. Alternatively, use Tencent Cloud Lighthouse for cloud deployment, which is completely unaffected by your local operating system.
Q4: How do I upgrade to the latest version?
A: Run the following command to automatically update:
hermes update
Q5: Approximately how much does the server cost?
A: Tencent Cloud Lighthouse's 2-core 4GB configuration package costs approximately $10–15 per month, with new user discounts available for first-time purchases. Compared to the cost of building and maintaining it yourself, this is a very worthwhile investment.
Congratulations! You now have your very own 24/7 cloud-based AI agent.
Let's review the entire process:
The true charm of Hermes Agent lies in its growth potential—the more you use it, the better it understands you and the more capable it becomes. From information retrieval to code writing, from file processing to automated operations, it will become an increasingly capable digital partner.
After deployment, you can also explore:
hermes dashboard to open the visual management panel to view conversation history, token statistics, and skill listshermes model to switch between different LLM providers at any timeIf you encounter any issues during deployment, feel free to leave a comment below. Happy herding! 🐴✨
Last updated: April 2026
Main references: Tencent Cloud Lighthouse official documentation, Hermes Agent GitHub project page, relevant technical reports from April 2026