To be fair, that comment is by Boris Cherny, the guy who developed Claude Code. You have to give him a bit of slack for wanting to use his tool for literally everything.

> They wont do it.

That's the point of this post, Boris Cherny closed it as "completed", of course without implementing it.

You're asking the wrong question here. There is:

How exactly does Claude Code* not support AGENTS.md?

How exactly does the Claude suite of Models not support AGENTS.md?

The problem being pointed at in parent linked is referring to the Claude Code Harness not supporting AGENTS.md. Harnesses which do support it (eg codex), append the content to the initial model turn upon the model discovering it at the project root.

* Or Claude Code Tui, Claude Code desktop, Claude Cowork, Claude Desktop, Claude Design, Etc.

(Edit: i wish I could tattoo the distinction on ny forehead. Im vocal about Anthropic engineer-oriented tools being lackluster, and i frequently find myself in conversations where I have to stop a coworker and ask if they're talking about a platform, tool, or model- and which, depending on the answer. I feel like those two things together make me come off somewhat abrasive, but man, we're all engineers here.)

Every other coding harness (i.e. 9 out of the top 10) automatically read AGENTS.md. It is a universal standard. Claude Code does not.

Claude code will only read the other .md files after you take a turn to prompt it, not when the session starts. So if I run `/clear` it will retain the CLAUDE.md context but not AGENTS.md.

Apple at least builds things with a modicum of quality control, which is not the case for Anthropic.

GitHub API says it was closed by bcherny himself:

$ gh api repos/anthropics/claude-code/issues/6235 --jq '{state, state_reason, closed_at, closed_by}'

{ "closed_at": "2026-08-17T03:37:37Z", "closed_by": { ... "id": 1761758, "login": "bcherny", "url": "https://api.github.com/users/bcherny", }, "state": "closed", "state_reason": "completed" }

I don’t see that the issue was closed by that PR. Where do you see that?

A good one does, yeah.

A bad one is noticeably harmful.

yes, in the tools I use, each (sub)agent will also load the same files (automatically, root is always loaded, a nested AGENTS.md is read if the dir or a peer file is touched)

yes, they are useful, mainly in that they shorten the context gathering phase and can call out gotchyas, keep it minimal

Anthropic are the only ones who think they can ignore the open standards. Absolute dirt bags.