AGENTS.md
A file in the environment that the harness loads into the context at session start: the project's standing brief to the agent.
An AGENTS.md file is a plain markdown file in the repository that the harness loads into the context window at session start. It is the project's standing brief to the agent: how to run the tests, what the conventions are, which directories are off limits, what the house style is.
The word standing is what makes it different from a prompt. A prompt applies to one request. AGENTS.md applies to every session in that directory, which makes it the right place for a rule that should never be forgotten: the build command, the lint rule people keep breaking, the file that must never be edited by hand.
It works because it is loaded, not because it is learned. The agent reads it the same way it reads any other file, so it occupies context and competes for attention, which means a long AGENTS.md is a real cost paid on every single call. The discipline is to keep it short and put detail behind pointers, letting the agent load the rest only when the task calls for it.
Related terms
Memory System
A system that makes an agent stateful across sessions by persisting to the environment and reloading at session start.
Context
The relevant information the agent has access to right now. A measure of quality, not quantity.
Progressive Disclosure
Show the model only the detail it needs right now, and load more when it asks.
Stateless
Carries no information forward. The model is stateless across requests; an agent is stateless across sessions by default.
Harness
Everything around the model that turns it into an agent: tools, system prompt, permissions, hooks.
Want this working inside your team?
It starts with one conversation about your workflow.