Pick Grok Bot if you want an agent working today without maintaining anything. Pick Hermes Agent if you want to choose the model, the tools and where it runs, and you're fine pasting commands into a terminal and fixing it when it breaks. Grok Bot is a managed product on its own cloud computer, from US$30 a month. Hermes is free, open-source software, so your bill is whatever model and hosting you plug into it.
I use Hermes with my ChatGPT subscription, and I run Grok Bot for jobs handled from my phone. This post is for the person already down to these two, and it's deliberately narrower than my four-agent comparison. Facts below are from the official docs, checked on 16/11/2026.
What's the difference between Hermes Agent and Grok Bot?
Grok Bot comes ready-made. Hermes comes as parts.
Grok Bot's bots share a persistent cloud computer with a browser, filesystem and terminal. You sign in, connect apps and start. Hermes is an open-source agent from Nous Research under an MIT licence. You install it, pick a model provider, choose where it runs, and wire up how you talk to it.
| Question | Grok Bot | Hermes Agent |
|---|---|---|
| Where it runs | One cloud computer shared by your bots | Your computer, or a server you pick (local, Docker, SSH, Modal, Daytona or Vercel Sandbox) |
| Setup | Sign in, connect apps | One install command, then choose model, tools and chat apps |
| Cost | From US$30 a month (Premium) | Software free. You pay for the model, plus hosting if it runs on a server. |
| Choose the model? | No, it picks per task | Yes, any provider you connect |
| Laptop shut | Routines keep running | Only if it runs on something that stays on |
Sources: xAI's Grok Bot FAQ and pricing, Hermes's GitHub page and install guide. Prices are US list prices before tax.
How much setup effort does each one take?
Grok Bot is the easier start. There's a desktop app and a phone app, and connectors for Gmail, Google Calendar, Google Drive, Slack, Notion and Granola. It also handles any MCP or API (the standard plug that lets an AI connect to another app). My Grok Bot setup guide is the order I'd work in.
Hermes's install is quick on paper. The docs say it's ready "in under two minutes" with one command on Mac, Linux or Windows (Mac and Linux installs need Git). The slow part comes after. You choose a model, then connect your tools yourself, which means MCPs, hunting down documentation and setting up API keys. That's the bit that stops most non-technical people, and it's where Grok Bot earns its price.
What do you pay for with Hermes?
Two things: the model and the hosting.
For the model, Hermes works with Nous Portal, OpenRouter, OpenAI or any endpoint. I use my ChatGPT subscription. Hermes's docs say its OpenAI Codex provider signs in with a device code, and that which ChatGPT plans are eligible isn't documented. Test with your own plan before you count on it.
Claude is trickier. Hermes's docs say the subscription route only works on a Claude Max plan with extra usage credits, and Pro subscribers can't use it. Or you use an API key, billed at standard API pricing.
For hosting, you only pay if you want it always on, and then you rent a server. Hermes's GitHub page suggests a $5 VPS as one option. Price it with the host you choose, because intro prices tend to renew higher.
So the cheaper-looking option can have the bigger bill. Grok Bot's one price covers the computer.
Which one is always on?
Grok Bot, with no effort. xAI says closing the app, laptop or phone doesn't stop a routine.
Hermes runs scheduled jobs through its gateway, and the gateway must be running. On a laptop that's asleep, jobs wait. Put it on a server and it's always on, but now you're maintaining a server. Hermes can message you on Telegram, Discord, Slack, WhatsApp and Signal, so your phone is covered either way.
Which gives you more control and privacy?
Hermes, if you set it up carefully. You decide where it runs. It checks commands against a list of dangerous patterns and asks you before running a match. Messaging access is denied by default until you list who may use it. Anything you send to the model still goes to that provider.
Grok Bot trades some control for convenience. All your bots share one computer, so xAI says not to treat separate bots as a security boundary. Training opt-out follows your Cursor privacy settings, and it needs cloud data storage. Check those before connecting client information.
How does Hermes compare with ChatGPT Dots?
Dots is the managed option from OpenAI, so it sits next to Grok Bot, not Hermes. OpenAI says your dot lives in the cloud with its own computer and browser, and it can hand work to Codex or ChatGPT Work. It's rolling out gradually to eligible Pro (US$100, US$200 and US$500) and Business Premium accounts, and Pro excludes the EEA, UK and Switzerland.
If you pay for Pro already, Dots is the managed pick. If you want your own parts, Hermes is. My comparison of all four is here.
A 15-minute checklist: which one are you?
Answer yes or no.
- Are you comfortable pasting a command into a terminal? No means Grok Bot.
- Do you want to pick the model, or run one you already pay for? Yes means Hermes.
- Must it run with your laptop shut and zero upkeep? Yes means Grok Bot.
- Will a team member use the same agent? Grok Bot's Team Bots are on Team and Enterprise plans.
- Are you willing to fix it when an update breaks it? No means Grok Bot.
- Does the work need to stay on a machine you control? Yes means Hermes.
Four or more answers pointing one way is your pick. A split means start with Grok Bot and revisit Hermes once you've written down a job worth automating.
Quick answers
What about OpenClaw?
Hermes's GitHub page lists a command, hermes import-agent openclaw, that brings across settings, memories and skills from OpenClaw, so people do move between them. This post sticks to Hermes and Grok Bot.
Can I run Hermes on my laptop?
Yes, it installs on Mac, Linux and Windows. Scheduled jobs only fire while the gateway is running, so a sleeping laptop pauses them.
Is Hermes safe to leave running?
Keep the defaults. The docs say never to use GATEWAY_ALLOW_ALL_USERS=true in production, and to set an explicit list of who can message it.
Get help with the setup
I run a paid AI membership, so I have an obvious interest here. Agents are where setups tend to break halfway, and that's when having someone to ask saves a weekend. Inside the membership, we set up and test agents live together on the monthly Build with Brooke, and the private community is there when yours stops working. Join the Wright Mode membership.


