AGENTS.md
AGENTS.md is a conventional document file placed in the root directory of open-source software projects. It is used to declare interaction rules, contribution norms, and operational constraints to AI coding assistants or automated agents. Its core function is to provide AI agents with project context guidance and specify the boundaries of permitted and prohibited operations, ensuring that automated tools conform to project maintainers' expected behavioral standards when participating in code generation, review, or modification.
Core Facts
Cursor uses the .cursorrules file to implement persistent preference functionality similar to CLAUDE.md.
AGENTS.md is a common naming convention that has emerged in the AI programming tools ecosystem, and tools such as OpenAI Codex and Cursor also recognize this file to varying degrees.
AGENTS.md is an emerging convention format used to define code style preferences, prohibited operation scopes, project-specific knowledge, and other content to guide agents in making decisions that conform to team norms during autonomous execution.
Timeline (last 90 days)
Codex reads `AGENTS.md` as its instruction file.
Recommended approach for mixed teams: keep `AGENTS.md` as the single source of truth and point to it with a single line in `claude.md`.
Codex and Cursor can directly read the `AGENTS.md` file.
The `AGENTS.md` standard does not define specific syntax, but rather convention on file location and basic structure (such as using Markdown headings to divide sections).
`AGENTS.md` is a community-driven open specification hosted by the Linux Foundation, aimed at providing a unified project instruction entry point for AI agents.
The CLAUDE.md/AGENTS.md file in the project repository contains two hard rules: PCSX2 must be credited, and no feature may be claimed to work without testing.
Based on anecdotal observation, the author believes Claude models are more willing than GPT models to follow conciseness requirements in system instruction files (such as CLAUDE.md and AGENTS.md), but this has not been quantified.
CLAUDE.md and AGENTS.md are both project-level instruction files, read into context at the start of a session by frameworks such as Claude Code and Codex respectively, but different harnesses handle instruction priority differently.
If the agent repeatedly ignores a rule, it is most likely because the file is too long and the rule is drowned out.
The Claude Code documentation recommends keeping each AGENTS.md file under 200 lines, because longer files consume more context and reduce adherence.
40 more timeline events
All Facts (20)
Cursor uses the .cursorrules file to implement persistent preference functionality similar to CLAUDE.md.
80%VerifiedAGENTS.md is a common naming convention that has emerged in the AI programming tools ecosystem, and tools such as OpenAI Codex and Cursor also recognize this file to varying degrees.
80%VerifiedAGENTS.md is an emerging convention format used to define code style preferences, prohibited operation scopes, project-specific knowledge, and other content to guide agents in making decisions that conform to team norms during autonomous execution.
75%VerifiedAGENTS.md是一种新兴实践，开发者通过在项目根目录放置Markdown文件向AI编码代理传达编码规范、架构偏好和工作流程要求
70%VerifiedClaude Code's official documentation explicitly states that it reads CLAUDE.md, not AGENTS.md.
65%VerifiedTypical AGENTS.md content includes a project architecture overview, a list of core files that must not be modified, test run commands, code style conventions, and change approval processes.
65%Verified执行 claude -v 可查看 Claude Code 版本，教程演示版本为 2.1.207
65%Verified优化上下文注入策略可能比更换更便宜的模型更能有效降低AI编程开支
65%VerifiedClaude Code v2.1.277 added support for the AGENTS.md configuration file.
80%UnverifiedAGENTS.md类似于传统的CONTRIBUTING.md，但专门面向AI代理而非人类贡献者
85%UnverifiedAGENTS.md的实践与'上下文工程'（Context Engineering）的理念密切相关，代表了从提示工程到上下文工程的演进
85%Unverified编码规则的价值不在于让AI写出超越自身能力的代码，而在于减少探索和决策过程中的浪费
85%UnverifiedAGENTS.md是用于描述业务逻辑上下文的配置文件，包含技术栈使用方式、文件组织映射和业务逻辑说明
85%UnverifiedAGENTS.md是一种新兴的项目级AI配置文件规范，最早由Google在其内部工程实践中推广
70%Partially VerifiedAPC is a portable project context layer, containing AGENTS.md, agent definitions, skills, and persistence rules that travel with the project.
65%Partially VerifiedWhen a CLAUDE.md file is not present in a folder, Claude automatically looks for and uses AGENTS.md as the source of project instructions.
65%UnverifiedThe concept of AGENTS.md was originally shaped by OpenAI Codex and community practices.
60%UnverifiedClaude Code's compatibility strategy is to read CLAUDE.md first and fall back to AGENTS.md when it is absent.
60%UnverifiedAn AI tool reading a context configuration file at startup is equivalent to injecting a project-level system prompt into the model before each conversation, thereby reducing low-quality modifications caused by a lack of background knowledge.
60%Unverifiedagents.md and CLAUDE.md are configuration files supported by Claude Code and similar tools; AI agents automatically read them at startup and inject them into the conversation context as system prompts.
60%