PulseFlow Tecnologia
← Insights
·6 min read·product-update·agents-md·agentic-ai-foundation·coding-agents
PulseFlow Tecnologia

AGENTS.md: the new configuration artifact for coding agents

A simple file format - markdown text, no rigid schema, no proprietary tool behind it - already has more than 22,000 GitHub stars, 1,700 forks, and was donated to a neutral foundation alongside the Model Context Protocol itself. AGENTS.md stopped being an informal convention some projects used and became the standard configuration artifact for coding agents.

A README, but for agents

The format's core proposal is simple: "a simple, open format for guiding coding agents" - one predictable place where a coding agent looks for instructions before touching a repository, covering development environment tips, testing instructions, PR guidelines, and project-specific commands and conventions. The README analogy isn't a stretch: just as a README orients a human developer joining a new project, AGENTS.md orients an agent.

What 2,500+ repositories already taught about the format

A GitHub analysis of more than 2,500 repositories with an AGENTS.md concluded that "successful agents aren't just vague helpers; they are specialists." Six blocks consistently show up in the most effective files: ready-to-copy executable commands, project structure, tech stack with specific versions, real code examples, style patterns, and clear boundaries on what the agent can and can't do alone. Among identified restrictions, the most common and most useful was simply "never commit secrets."

From project convention to a foundation's founding project

The most important maturity leap wasn't technical - it was governance. AGENTS.md, alongside Anthropic's Model Context Protocol and Block's goose, is a founding project of the newly created Agentic AI Foundation, under the Linux Foundation umbrella, with Amazon Web Services, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI as Platinum members. A file format that, until recently, was maintained independently now has the same neutral governance structure backing critical open source infrastructure projects.

Why this matters beyond convenience

An AGENTS.md versioned alongside the code solves a problem external documentation never really solved: staying current. Documentation in a wiki or knowledge tool separate from the repository tends to go stale silently; a file inside the repository itself, reviewed in the same pull request that changes the architecture it describes, has a much better chance of staying accurate.

What this signals for anyone who hasn't adopted it yet

With adoption measured in tens of thousands of repositories, an empirically tested six-block pattern, and now shared governance among the world's largest AI companies, the question for engineering teams stopped being "is it worth having an AGENTS.md?" and became "why doesn't our repository have one yet."

Sources