Claude Code 最大の要望 AGENTS.md 対応——5,200 を超える reactions の痛みと今すぐできる回避策 Claude Code 最大の要望 AGENTS.md 対応——5,200 を超える reactions の痛みと今すぐできる回避策

In the world of AI coding agents, how you pass project-specific context and rules to the agent matters enormously. One format that has emerged as a de facto standard is the AGENTS.md file — a plain Markdown file placed at the repository root that tells an agent how to behave within that project. OpenAI's Codex and several other tools adopted this convention early, and it has since become a common expectation among developer teams juggling multiple AI tools.

Claude Code, Anthropic's terminal-native coding agent, does not currently support AGENTS.md. The gap has become the most loudly voiced complaint in the tool's GitHub repository: a single issue requesting AGENTS.md support has accumulated 5,196 reactions, making it the highest-voted open item by a significant margin. That number signals not just frustration but a real barrier to adoption, particularly for teams that already have AGENTS.md files baked into their workflows.

Claude Code does have its own equivalent — CLAUDE.md — which serves a similar purpose of injecting project-level instructions. The problem is that the two formats are not interchangeable. Teams that use Claude Code alongside other agents like Cursor or Copilot must maintain parallel files, duplicating effort and risking drift between the two sources of truth. For organizations trying to standardize AI tooling across large engineering teams, this kind of friction is often enough to tip the scales toward a competing product.

The article proposes five workarounds to bridge the gap today. The most straightforward is simply copying the contents of AGENTS.md into CLAUDE.md manually, which works but requires ongoing synchronization. A more elegant solution is using symbolic links so that both files point to the same underlying content, minimizing duplication while keeping Git history clean. CI-based auto-generation is another option for teams with mature pipelines, where a script maintains both files from a single source of truth. Each approach has trade-offs in complexity and maintenance burden, but all are immediately actionable.

Zooming out, the AGENTS.md situation reflects a broader fragmentation problem in the AI developer tooling ecosystem. Every major player — Anthropic, Microsoft, OpenAI, and others — has introduced slightly different conventions for project-level context files. Until a genuine cross-tool standard emerges, developers will continue paying the tax of managing multiple formats. Whether AGENTS.md ultimately wins out as the industry norm remains to be seen, but its adoption curve is pointing upward.

For Anthropic, the pressure is hard to ignore. Five thousand-plus reactions on a single GitHub issue is an unusually strong signal, and the longer official support is delayed, the more Claude Code risks ceding ground to competitors in team and enterprise settings where standardization is non-negotiable. No public timeline has been announced as of this writing, but the community weight behind the request suggests it is unlikely to stay on the back burner indefinitely.