Feature Request: Support AGENTS.md

A feature request for Claude Code to support the open-standard AGENTS.md alongside its proprietary CLAUDE.md has ignited a fiery debate on Hacker News. Users are lambasting Anthropic's perceived corporate arrogance and a move towards vendor lock-in, drawing parallels to platforms like Twitter and Reddit that alienated their communities. The discussion highlights broader frustrations with proprietary AI tools and calls out the growing trend of 'enshittification' in the tech industry.

Score

Comments

Highest Rank

12h

on Front Page

First Seen

Aug 19, 11:00 PM

Last Seen

Aug 20, 10:00 AM

Rank Over Time

The Lowdown

The core of this Hacker News post is a feature request for Anthropic's Claude Code, an AI coding assistant, to adopt a dual-file approach for project context by supporting the widely-used AGENTS.md open standard as a fallback to its native CLAUDE.md.

Currently, Claude Code exclusively uses CLAUDE.md for project-specific instructions, ignoring any AGENTS.md files that might be present.
AGENTS.md is described as a gaining traction open standard, utilized by over 20,000 open-source projects for general agent-agnostic context (e.g., setup commands, testing instructions).
The proposal suggests a lookup logic: Claude Code would first check for CLAUDE.md, and if not found, then look for AGENTS.md.
This change aims to improve interoperability, lower the barrier to using Claude Code on existing open-source projects, and ensure backward compatibility without compromising Claude's specific features.
Benefits include seamless onboarding for developers, better team interoperability where multiple AI agents are used, and no breaking changes for existing CLAUDE.md setups.

By embracing this open standard, the request argues Claude Code could become a more collaborative tool and better integrate into the broader AI development ecosystem, reducing friction for its users.

The Gossip

Corporate Calculus and Community Chasm

Many commenters see Anthropic's refusal to support `AGENTS.md` as a deliberate strategy to promote its proprietary `CLAUDE.md`, akin to "free advertisement" or building a "moat." They draw parallels to historical instances where platforms like Reddit and Twitter alienated users by closing off open standards and third-party tools, leading to "enshittification" and a decline in quality content or user engagement. This decision is viewed as a sign of institutional arrogance and a drift away from user needs towards internal corporate interests.

Annoying Adherence and Absurd Workarounds

Users highlight the practical friction caused by Claude Code's inflexibility, pointing out the necessity of creating redundant `CLAUDE.md` files or symlinks just to utilize the existing `AGENTS.md` context. This is seen as an unnecessary and easily fixable overhead, with several commenters remarking on the trivial technical effort it would take for Anthropic to implement the proposed dual-file support. The discussion also touches on the limitations of these workarounds for advanced features like lazy loading in subdirectories, and one user humorously dismisses concerns about symlink storage cost.

Broader Discontent with Claude Code's Direction

Beyond the `AGENTS.md` issue, some commenters express general disillusionment with Claude Code. They criticize its perceived decline in quality, describing it as an "anchor" that is getting worse, with essential information hidden behind "bloated useless TUI" and "rambling tangential responses." The perceived market leadership of OpenAI and Codex is brought up, suggesting Anthropic's product decisions are making Claude Code a less competitive and attractive option.