# AGENTS.md

> **TL;DR.** AGENTS.

- **Category:** AI / Developer Tools / Conventions
- **Stage:** established
- **Age:** 287 days
- **Origin date:** 2025-08-20
- **First detected:** 2026-04-15
- **Canonical URL:** https://earlyterms.com/term/agents-md
- **Sources:** 10 primary URLs

## Definition

AGENTS.md is an open, vendor-neutral markdown file placed at the root of a repository that tells AI coding agents (Claude Code, Codex CLI, Cursor, Copilot, Jules, Amp, Factory, Windsurf) how to work with the codebase. It is to AI agents what README.md is to humans — build commands, conventions, boundaries, in plain markdown with no custom syntax.

The convention crystallized between August 2025 and early 2026 through a multi-vendor collaboration (OpenAI, Google, Sourcegraph, Cursor, Factory) and is now stewarded by the [Agentic AI Foundation](https://agents.md/) under the Linux Foundation. The [initial Hacker News launch](https://news.ycombinator.com/item?id=44957443) hit 837 points on Aug 20, 2025. By April 2026 the spec site reports adoption by 60,000+ open-source projects and the reference repo carries 20.3k stars.

## Example

GitHub's Copilot team [studied 2,500+ AGENTS.md files](https://github.blog/ai-and-ml/github-copilot/how-to-write-a-great-agents-md-lessons-from-over-2500-repositories/) in Nov 2025 and distilled a six-component template: executable build commands early, real code snippets, explicit `never modify` boundaries, exact stack versions, a three-tier permission model, and a concrete agent persona.

## Analogy

README.md for AI agents — same root-of-the-repo slot, same markdown grammar, different reader.

## Why it's emerging now

By Apr 2026, AGENTS.md has Linux Foundation stewardship, 60k+ adopting repos, and a real research literature — Vercel's 33/33 vs 31/33 eval result and a Feb arXiv paper gave the convention evidence-backed credibility. Anthropic's Claude Code remains the one major holdout, with feature request #6235 among its most-reacted issues.

## Related terms

- *alias:* CLAUDE.md
- *alias:* .cursorrules
- *alias:* .github/copilot-instructions.md
- *related:* README.md
- *related:* Claude Code
- *related:* OpenAI Codex CLI
- *related:* Cursor
- *related:* GitHub Copilot
- *parent:* Coding Agents
- *competitor:* Agent Skills
- *related:* MCP Server
- *related:* agent harness
- *related:* Agentic AI Foundation

## Sources

1. [agents.md — official specification site](https://agents.md/)
2. [agentsmd/agents.md on GitHub (20.3k stars)](https://github.com/agentsmd/agents.md)
3. [Hacker News — AGENTS.md: Open format for guiding coding agents (837 pts, Aug 20, 2025)](https://news.ycombinator.com/item?id=44957443)
4. [InfoQ — AGENTS.md Emerges as Open Standard for AI Coding Agents](https://www.infoq.com/news/2025/08/agents-md/)
5. [GitHub Blog — How to write a great agents.md (Nov 2025, 2,500-repo study)](https://github.blog/ai-and-ml/github-copilot/how-to-write-a-great-agents-md-lessons-from-over-2500-repositories/)
6. [Vercel — AGENTS.md outperforms skills in our agent evals (524 HN pts, Jan 29, 2026)](https://vercel.com/blog/agents-md-outperforms-skills-in-our-agent-evals)
7. [OpenAI Codex — Custom instructions with AGENTS.md](https://developers.openai.com/codex/guides/agents-md)
8. [Anthropic Claude Code — Feature Request: Support AGENTS.md (#6235)](https://github.com/anthropics/claude-code/issues/6235)
9. [The Prompt Shelf — AGENTS.md vs CLAUDE.md: What's the Difference](https://thepromptshelf.dev/blog/agents-md-vs-claude-md/)
10. [arXiv — Evaluating AGENTS.md: Are They Helpful for Coding Agents? (232 HN pts)](https://arxiv.org/abs/2602.11988)

---
_Generated by EarlyTerms · https://earlyterms.com/term/agents-md_