How Does agents.md Compare to claude.md, gemini.md, and Cursor Rules, and Which Approach Is Becoming the De Facto Standard?

For a while, every AI coding tool wanted its own way of being told about your project. Anthropic had claude.md, Google leaned toward gemini.md, Cursor had its rules files, and others added their own variants. They all did the same job, briefing an agent on your codebase, but in incompatible ways, which meant maintaining several files that said the same thing. That fragmentation is now resolving, and understanding how these formats compare tells you where to place your bets.

They all solve the same problem

Despite the different names, these files are variations on one idea. Each is a markdown document that gives a coding agent the project context it cannot infer, exactly as the concept of an agents.md file describes. Build commands, conventions, architecture, and warnings go in, and the agent reads them before working. The differences are in naming and which tool reads which file, not in the underlying purpose. At heart, they are the same tool wearing different labels.

claude.md was the pioneer

The concept was popularized by one vendor before it spread. Anthropic’s Claude Code introduced claude.md as a place to hand the agent durable project context, and it caught on fast because it so obviously worked. Its success is what proved the whole idea that a simple markdown briefing dramatically improves agent behavior. In that sense, claude.md is the ancestor of everything that followed. It demonstrated the value that later drove standardization.

Then everyone rolled their own

Success bred imitation, and imitation bred fragmentation. Other tools adopted the pattern under their own names and formats, so a team using several tools ended up maintaining claude.md, a Cursor rules file, and more, all describing the same project. This duplication is precisely the kind of overhead that good practice tries to avoid, since it creates several places for the truth to drift out of sync. The idea had won, but the lack of a shared format made it needlessly painful. Everyone agreed on the concept and disagreed on the file.

agents.md became the open standard

The resolution came through a shared specification rather than one vendor winning. The AGENTS.md standard was formalized as an open spec in 2025, led by OpenAI with participation from Google, Cursor, and others, and later donated to a neutral foundation. By late 2025, tens of thousands of open-source projects had adopted it and dozens of tools supported reading it. A single file could now brief many different agents, which is exactly what the fragmented approach could not do. An open standard succeeded where competing formats could not.

Why a shared standard matters

The value of convergence is not abstract, it is one less file to maintain and one less way to drift. When rival companies collaborate on a format, as they did through the Agentic AI Foundation, they are agreeing that the format is shared infrastructure rather than a competitive edge. That kind of cooperation is a strong signal that the standard will stick, because everyone has an interest in it working. Standards backed by the whole industry tend to outlast any single tool. Shared plumbing is more durable than proprietary plumbing.

The de facto winner

Putting the pieces together, agents.md is emerging as the default. It has the open specification, the neutral governance, the broad tool support, and the huge base of projects already using it, which together are what make a standard de facto rather than merely proposed. The vendor-specific files still exist, but the momentum has clearly shifted to the shared format. When the industry converges like this, betting against the standard is usually a mistake. The direction of travel is unmistakable.

What about the vendor files

The older files have not vanished, and you do not have to abandon them overnight. Many tools still read their own format, and some read both their native file and agents.md, so a transition can be gradual. A common bridge is to keep a single source of truth in agents.md and point the vendor-specific file at it, avoiding duplicate content. This lets you adopt the standard without breaking tools that have not fully caught up. Compatibility during the shift is manageable rather than painful.

The practical recommendation

For a new project, the simple advice is to start with agents.md and treat it as your primary context file. It is the format most likely to be read by whatever tools you use now and adopt later, which protects your effort from tool churn. Because it is genuine context engineering, the care you put into it pays off regardless of which agent reads it. Investing in the standard is investing in something that will still matter after your current tool is replaced. Bet on the format the whole field is converging on.

Content matters more than format

It is worth keeping perspective on what actually drives results. The format is a container, and the value is in the context you put inside it, so a great claude.md beats an empty agents.md every time. Whichever file you use, the real work is the same act of context engineering, choosing the high-signal guidance an agent needs. Do not let the standards debate distract from writing good context in the first place. The wrapper matters far less than what is written on the label.

Watch the convergence, not the noise

This whole episode is a clean example of a broader pattern worth learning. When an industry moves from many competing formats to one shared standard, that convergence is a much stronger signal than any individual product announcement. Following the standards, rather than the launches, is a reliable way to cut through the hype and spend your effort well. The formats that get standardized are the ones that will still be here in a few years. Convergence, not marketing, tells you what is real.

The takeaway

agents.md, claude.md, gemini.md, and Cursor rules all do the same job of briefing a coding agent, but agents.md has become the de facto standard through an open specification, neutral governance, and broad adoption. Start new projects with agents.md, bridge older files to it gradually, and remember that the content matters more than the container. Bet on the shared standard, and write good context to put in it.

Common questions