Hermes AI Agent¶
Hermes AI Agent runs Hermes Agent, the open-source AI agent from Nous Research, on your own EC2 instance, and lets you talk to it through Telegram. The image comes with a pinned Hermes release on hardened Ubuntu 24.04. You add your model-provider key and Telegram bot after launch with one script.
Launch on AWS Marketplace Getting started
What you get¶
- Ubuntu 24.04 LTS with a 20 GB gp3 root volume and a 4 GB swap file.
- Hermes Agent, pinned to the release shown in the listing's version title, installed in its own Python environment.
- Messaging adapters for Telegram, Discord and Slack, and the native Anthropic SDK. First-boot setup configures Telegram; you can add the others yourself.
- A provisioning script,
/opt/hermes/bin/provision-hermes.sh, that writes your settings and starts thehermes-gatewayservice. - Security hardening that is on from first boot: a UFW firewall, fail2ban, key-only SSH, kernel network hardening, daily security updates and required IMDSv2. See Operations.
No credentials are baked into the image. Browser automation is not installed.
Supported model providers: OpenAI, Anthropic, OpenRouter, Google Gemini and Groq. You bring your own API key.
Ports¶
| Port | Protocol | Used for |
|---|---|---|
| 22 | TCP | SSH, as the ubuntu user, with your EC2 key pair. |
That is the only inbound port. Telegram and your model provider are reached over outbound HTTPS, so the instance needs outbound internet access.
Recommended instance¶
A t3.medium. The listing also offers t3.small, t3.large, t3.xlarge,
m6i.large and m6i.xlarge.
Billing¶
- The software costs USD 0.02 per running instance-hour.
- A five-day free trial covers the software for one instance. When it ends it turns into the paid hourly offer automatically.
- AWS bills EC2, EBS and data transfer separately, trial or not. Your model provider bills its own usage.
- Stop or terminate the instance to stop the hourly charge. Use Manage subscriptions in AWS Marketplace to review or cancel the subscription.
Where to go next¶
- Getting started: launch, connect and provision your agent.
- Configuration: settings, models and channels.
- Operations: security, logs, backups and upgrades.
- Troubleshooting: fixes for common problems.
Support¶
For help with the image, deployment or billing, email info@softwaresushi.com. Support is in English during US business hours. Report security issues to the same address with the subject Hermes AI Agent security.
Upstream project¶
Hermes Agent is developed by Nous Research and released under the MIT license. Software Sushi packages it for AWS and does not claim authorship. For the agent itself, see the Hermes Agent repository and its issue tracker.