About two weeks ago, I went to Reddit with a question: "How do we call software like OpenClaw or Hermes — AI agentic software designed to work on its own without human intervention?" I was surprised to find that there is no established term for it. I expected something like "AI agentic software" or "AI agentic system" to already exist, but there isn't a clear standard.
The discussion under my post was short but interesting. About half the participants mentioned the term "AI harness" — though not always in exactly that form. Variations like "AI harnessing software" or "AI agent harness" came up too. The other half disagreed entirely, arguing the term is incorrect and that "AI agentic software" or "AI agentic system" should be used instead.
Furthermore, I found that some people use "AI harness" for other kinds of software as well — tools like Codex or Cursor, for example. To me, those belong in a different category: AI coding agents. That's not the same thing.
Since then, I've started noticing the term "AI harness" — or sometimes just "harness" — appearing more and more in posts and discussions. My impression is that it's going to win and become the standard term for this class of software. Whether it will be used specifically for "AI agents with initiative" — systems that act without being prompted — that question is still open.
We do need a clear term for LLM-based agentic software that operates autonomously in a loop without constant human guidance. AI coding agents don't work on their own initiative, so they shouldn't be lumped into the same group. Separate group and term are needed for tools like OpenClaw, Hermes, DMJBot and others that are designed to operate independently. I think "AI harness" is a good candidate for that term.
I have a feeling that over the next few months we'll see this term settle into common use. And when it does, it will be a useful shorthand for a genuinely new category of software.
