Claude Code now reads AGENTS.md: one instruction file for every AI coding tool
TL;DR
Teams running several AI coding tools often maintain several config files (CLAUDE.md, .cursor/rules, AGENTS.md…). Claude Code now reads AGENTS.md when a folder has no CLAUDE.md, so one instruction file can serve multiple tools.
What changed
| Point | Details |
|---|---|
| Version | Claude Code 2.1.277 |
| Precedence | CLAUDE.md wins; AGENTS.md is used only in folders without a CLAUDE.md |
| Configurable | Use /config to read AGENTS.md, CLAUDE.md, or both |
| Exception | Requires the session to fetch feature flags from Anthropic; via Bedrock/other providers or with telemetry off it falls back to CLAUDE.md |
| AGENTS.md | Open, plain-Markdown format at the repo root, governed by the Agentic AI Foundation (Linux Foundation) |
How to put it to work
- Multi-tool repos: move shared guidance (build, test, code conventions) into AGENTS.md.
- Keep CLAUDE.md thin for Claude-only features (hooks, skills, subagents) and point to AGENTS.md for the rest.
- Verify per tool: the same instructions may be interpreted differently by different agents.
- Tech Leads: treat AGENTS.md like onboarding docs — written for humans and machines.
Further reading: InfoWorld