Looking for a Hermes Alternative? Keep the AI You Already Use

You may not need a new agent at all. SentinelX connects your servers to Claude or ChatGPT, so the assistant you already use can act on them, safely.

Hermes has put a question on a lot of people's minds this year: what if my AI could actually do things on my machines, instead of just telling me what to type?

It's a good question. But before you adopt a new agent to answer it, it's worth asking a second one: do you need a new agent at all?

You already have an assistant you trust

Most people reading this already use Claude or ChatGPT every day. You know how they think, you've built up conversations, projects and habits around them, and they're on your laptop and your phone.

What they're missing isn't intelligence. It's reach. They can explain exactly how to check why a service is down, but they can't check it. They can write the fix, but you're the one who has to paste it into a terminal.

SentinelX closes that gap without replacing anything. It connects your servers to Claude or ChatGPT, so the assistant you already talk to can act on them. It doesn't bring a new brain. It gives the one you already have hands.

How it works

There are two pieces.

A small agent runs on each machine you want to manage, whether that's a VPS, a home server, a laptop or a Windows box. And a connector plugs SentinelX into Claude or ChatGPT, the same way you'd add any other connector. That's the bridge: your server on one side, your AI on the other.

After that you just ask Claude or ChatGPT, in plain language:

"Why is nginx returning 502 on the web server?"

"Free up some disk on the backup box."

"Show me what changed in the config since yesterday."

Your assistant reads logs, runs commands, edits files and reports back, in the same chat you were already using.

Simple to set up

Installing the agent is one line:

curl -fsSL https://get.sentinelx.app | bash

It runs on Linux, macOS and Windows. It connects out to SentinelX over a single WebSocket, so there are no inbound ports to open, no firewall rules to write and nothing exposed to the internet. A machine behind NAT at home works exactly like a cloud VM.

And because it works through Claude and ChatGPT, it goes wherever they do, including their mobile apps. You can check on a server from your phone.

Safe by design

Giving an AI access to a real machine is exactly where you want to be careful, and SentinelX is built around that.

The host decides what's allowed, not the AI. Every agent has an allowlist that you control, and anything not on it is refused. It's deny by default. The assistant can't talk its way past it or grant itself more access, because the policy lives on the machine, not in the conversation.

Everything is audited. Every command, every file read, every edit is recorded. You can see what your assistant did, on which host, and when.

It starts conservative. The agent doesn't get elevated privileges unless you explicitly give them, which means the cautious setup is the default one.

What you keep

This is the part that's easy to overlook. With SentinelX you don't move your work somewhere new. You keep your assistant, your history, your projects, and the way you already talk to it. You're not learning a new tool or a new place to chat.

You're just letting the AI you already rely on reach the machines you already run.

Getting started

Install the agent on a machine, add the SentinelX connector to Claude or ChatGPT, and ask it something about that host. The first time it answers with something it actually looked up, rather than something it guessed, the difference is obvious.

Details and plans at sentinelx.app.