Hermes Agent from Nous Research is a strong open agent. It keeps a memory, writes its own skills, runs scheduled jobs and answers in Telegram or Slack. It is free under MIT. To set it up, you run an install script, then edit ~/.hermes/config.yaml and put your keys in ~/.hermes/.env.
Off Grid AI gives a normal person the same kind of help without the setup. It is a finished app that runs your own models, already knows your work, and waits for your yes before it acts.
Get Pro for $79.99 once · Download Off Grid AI
Off Grid AI and Hermes Agent side by side
| Off Grid AI | Hermes Agent | |
|---|---|---|
| How you install it | Download the app | Install script, or desktop app on macOS 12+ and Windows 10/11 |
| Where settings live | In the app |
~/.hermes/config.yaml, keys in ~/.hermes/.env
|
| Models | Local models you download in the app | Your own provider, Nous Portal, or a self-hosted server |
| Before it acts | Mail, issues and docs wait for your approval (Pro) | Smart mode: a second model approves low-risk commands |
| Memory | Your screens, meetings, mail and docs, on your disk (Pro) | Conversations, memories and skills it saves |
| Scheduled work | God routines and scheduled tasks (Pro) | Natural-language scheduling through its gateway |
| Phone | iPhone and Android apps with their own models | Messaging apps; Android through Termux |
| Price | Free app; Pro $79.99 once or $7.99 a month | Free; Portal $0, $20, $100 or $200 a month |
Checked 30 October 2026 against the Hermes Agent site, its docs, configuration and security pages. Off Grid AI's Pro price is a limited-time lifetime offer, down from $119.
What setup looks like
The Hermes site gives a curl install script for macOS, Linux and WSL2, a PowerShell script for Windows, and desktop downloads for macOS 12+ and Windows 10/11. Linux uses the terminal install.
The docs recommend hermes setup --portal after install. That signs you in to Nous Portal for a model and its hosted tools. If you want a different provider, the configuration page says settings go in config.yaml and secrets go in .env.
Local models work. The provider docs list self-hosted endpoints such as Ollama and vLLM, and LM Studio. You install and run that server, then point Hermes at it.
In Off Grid AI Desktop (OGAD), you open Models and download one. The app shows which models fit your computer first. That is the whole setup.
What runs without asking
Hermes has a real approval system, and its security page is clear about it:
- The default for dangerous commands is
smartmode. A second model rates the risk. It auto-approves low-risk commands, denies clearly dangerous ones and asks you about the rest. - On container backends such as Docker or Modal, dangerous-command checks are skipped, because the container is the boundary.
- YOLO mode bypasses all approval prompts for the session. A hard blocklist still applies.
- Cron jobs and webhook sessions deny dangerous commands by default, unless you change them.
This is a good design for developers. You still have to understand it, and a model decides what counts as low risk.
Off Grid AI keeps the decision with you. With Pro, Actions drafts replies, issues and docs through Gmail, Linear, Jira and Notion. Each one waits in an approval card with the full text. Nothing goes out until you press Approve. God follows rules you set, such as "never send anything to Acme without asking".
A chief of staff that already knows your day
Hermes remembers what happens in its own sessions. That is useful, but you have to bring the context to it.
Off Grid AI Pro sees your day. Your screens, meetings, mail and documents become one local memory. God, your chief of staff, works from that memory:
- A morning briefing with the approvals that wait for you.
- Meeting prep with notes from the last meeting.
- Routines and scheduled tasks, such as a weekday brief.
- What people owe you, and what you owe them.
Meetings are recorded without a bot and transcribed on your computer. Reflect shows where your time went, with no timers.
What it costs
Hermes Agent is free. Nous Portal is optional. Its Free tier is "free models only". Plus is $20 a month, Super $100 and Ultra $200, each with monthly credits. If you use your own provider, you pay that provider instead.
| 1 year | 3 years | |
|---|---|---|
| Nous Portal Plus | $240 | $720 |
| Nous Portal Super | $1,200 | $3,600 |
| Off Grid AI Pro, lifetime | $79.99 | $79.99 |
With local models, Off Grid AI has no usage bill. The model runs on hardware you already own.
When Hermes Agent is the better pick
Be fair to Hermes. It is free, MIT licensed and very flexible. It can write and reuse its own skills. It spawns subagents with their own terminals. It runs on five backends: local, Docker, SSH, Singularity and Modal. It talks to you in Telegram, Discord, Slack, WhatsApp, Signal and email. It shipped v0.22.5 on 23 October 2026, per its GitHub releases.
If you are a developer who wants a scriptable agent on a server, Hermes is a fine choice. Off Grid AI does not live in your chat apps or run on a remote box.
If you want an assistant that is ready in ten minutes, sees your work, runs on your own models and asks first, use Off Grid AI.
Switch in ten minutes
- Download OGAD for Mac (Apple silicon), Windows or Linux.
- In Models, download a text model that fits your computer.
- Get Pro and enter your key in the app.
- Connect Gmail, Linear, Jira or Notion in Integrations.
- Open God, set a weekday briefing and read the first one.
Your MCP servers can come too. See how to connect a local AI to your apps through MCP.
FAQ
Is Hermes Agent free?
Yes. Hermes Agent is free and open source under the MIT license. Nous Portal, its optional model and tool service, has a Free tier with free models only, and paid tiers from $20 a month.
Does Hermes Agent ask before it runs commands?
Its security docs say the default is smart mode. A second model auto-approves low-risk commands, denies clearly dangerous ones and asks you about the rest. On container backends, the dangerous-command checks are skipped.
Can Hermes Agent use local models?
Yes. Its provider docs list self-hosted endpoints such as Ollama and vLLM, and LM Studio. You run that server yourself and set it in the Hermes config.
How is Off Grid AI different from Hermes Agent?
Off Grid AI is a finished app. Models download inside it, with no config file or API key. Pro builds a local memory from your screens, meetings, mail and docs, and its actions in Gmail, Linear, Jira and Notion wait for your approval.
What does Off Grid AI Pro cost?
On 30 October 2026, Pro costs $79.99 once for lifetime access, a limited-time price down from $119, or $7.99 a month. One key covers up to 5 devices.
Get Off Grid AI Pro and get a chief of staff that asks first.
Top comments (0)