Who builds Agent Friendly Code, and why
Who
Built and maintained by Himanshu Singh , a software engineer who builds side projects in the open. X: @hsnice16 .
Independent: not tied to Anthropic, OpenAI, Google, Cognition, Anysphere, or any other company whose agent is scored here.
Why it exists
A README alone doesn't help an AI coding agent get work done, and nothing public showed which repos do more. Agent Friendly Code shows it per agent, because agents read different files: Claude Code loads CLAUDE.md, Cursor its own rules and AGENTS.md, and Devin needs a dev setup it can run.
What it is not
Not a test of how well agents perform. Scores come from file checks, and no agent is run. Weights follow each agent's docs but aren't yet measured against real runs. The methodology has the plan to measure them.
How repos are chosen
The leaderboard is a curated list of 401 public repos on GitHub, GitLab and Bitbucket, kept in scripts/seed-list.ts and grouped by ecosystem: JavaScript and TypeScript, Python, Rust, Go, the JVM, other languages, and AI coding agents, models and SDKs. It is rescored every six hours; a repo taken off the list leaves the leaderboard at the next rescore.
A weekly audit flags repos on the list that were renamed, archived, deleted or made private, that are forks or mirrors, or whose score stopped refreshing; each is fixed or accepted on purpose. Scoring a repo with Live Score never adds it. To suggest one, open an issue or a pull request that edits the list.
This site's own repo, hsnice16/agent-friendly-code, and the author's other repos are on the list and scored by the same rules. They are labelled “this site's own repo” and “by this site's author” wherever repos are listed.
Open source
MIT-licensed. The checks, weights, scoring code, repo list and every score are in the source repository . If a score looks wrong, open an issue with a link and the rule in question. Missing a check? Suggest one.
Contact
Open an issue or discussion on GitHub . For anything private, email hsnice16@gmail.com.