We just published LLMS.md, an open specification for helping Large Language Models (LLMs) understand code repositories.
The idea is simple: • README.md helps humans • robots.txt helps crawlers • LLMS.md helps AI coding agents (Copilot, Cursor, Claude Code, Gemini CLI, etc.)
A root-level LLMS.md gives agents: • Key directories & entry points • Which files to ignore (dist, deps) • Run/test instructions • Reasoning hints from maintainers
Spec: github.com/llmspec/llms-spec RFC: llmspec/llms-spec/rfc/0001-llms-md-spec.md
We’d love feedback, contributors, and early adopters. If you maintain an OSS repo, try adding a LLMS.md and let us know what worked!