Anthropic adds support for the AGENTS.md instructions spec to Claude Code; OpenAI contributed AGENTS.md to the Agentic AI Foundation last year
Devs cheer relief from compatibility headache — Anthropic on Friday surprised the developer community by supporting rival OpenAI's mechanism for passing marching orders to AI agents.
Context & Ripple Effects
Anthropic has been building Claude Code from reusable instruction bundles in Skills for Claude toward coordinated agent teams and a managed-agent harness. That makes the format carrying project-level instructions a practical compatibility issue rather than a cosmetic developer preference.
OpenAI's contribution of AGENTS.md to the Agentic AI Foundation gives the format a standards venue. Anthropic's implementation is deliberately conditional: Claude Code checks AGENTS.md when a folder lacks CLAUDE.md, preserving its own file as the higher-priority configuration.
First-order effects
- Teams sharing repositories across coding agents can use AGENTS.md as a fallback in Claude Code instead of maintaining a separate CLAUDE.md solely for Claude compatibility.
- Anthropic gains a standards-compatible entry point for Claude Code while retaining CLAUDE.md as its native override and tying the support to its forthcoming Mods customization system.
Second-order effects
- OpenAI's AGENTS.md specification becomes more useful to developers running mixed-agent workflows because a major competing coding agent can consume the same repository instructions.
- Repository owners can consolidate baseline agent guidance, shifting differentiation among coding-agent vendors toward harness behavior, customization, and execution quality rather than duplicate instruction files.
Third-order effects
- If other agent vendors adopt the foundation-backed format, project instructions can become a portable layer across agent harnesses, reducing format lock-in while leaving each vendor control over precedence and extensions.
- The fallback design signals that interoperability may emerge through common baseline files rather than identical configurations, with vendors preserving proprietary control points above the shared standard.
The trend: Agent developers are standardizing portable repository instructions while competing over the harnesses and customization layers that interpret them.