React Doctor v2 launches npx react-doctor for Next.js, Vite, and React Native

• aidenybai's launch post introduced React Doctor v2 as an open source CLI aimed squarely at AI-written React code, with support for Next.js, Vite, and React Native.
• On the official React Doctor site, the tool is framed as a codebase health scanner that outputs a 0 to 100 score plus actionable diagnostics, while aidenybai's follow-up points to the public repo and the npx react-doctor@latest entrypoint.
• The current docs list flags for diff-only scans, JSON output, GitHub Actions annotations, staged-file checks, and agent installs for more than 50 coding tools, according to the homepage and GitHub repo.
• aidenybai's leaderboard post also shipped a public leaderboard that ranks open source React projects, with nodejs.org at 88 and tldraw plus Excalidraw at 84.

You can browse the repo, scan a project with the homepage command reference, and check the public leaderboard that shipped alongside aidenybai's leaderboard post. The odd but useful detail is that the docs already treat coding agents as a first-class surface, with a dedicated install command and support for agents including Pi, Cursor, Codex, and Claude Code on the official site.

The core pitch is simple: run npx react-doctor@latest and get a React health score instead of manually hunting through AI-generated diffs. aidenybai's launch post says it works across Next.js, Vite, and React Native, while the repo README describes a 0 to 100 score with framework-aware rules.

The homepage documents a slightly richer CLI than the launch post shows:

• --diff scans only changes against a base branch.
• --staged limits checks to staged files for pre-commit hooks.
• --json emits a structured report.
• --annotations turns findings into GitHub Actions annotations.
• --fail-on controls whether warnings or errors fail CI.
• --offline skips the score API and computes locally.

React Doctor is being sold as agent guardrails as much as a linter. The official docs include npx -y react-doctor@latest install, which adds the tool to coding agents and names support for more than 50 agent surfaces.

The GitHub repo says the scan reports issues across:

• state and effects
• performance
• architecture
• security
• accessibility
• dead code

That combination is broader than a style checker. It is closer to a framework-aware audit pass designed to sit after code generation.

The companion leaderboard gives the launch a benchmark-y hook instead of just a CLI demo. According to the official page, the first published rankings are:

1. nodejs.org, 88/100
2. tldraw, 84/100
3. Excalidraw, 84/100
4. Twenty, 78/100
5. Plane, 78/100
6. Formbricks, 75/100
7. PostHog, 72/100
8. Supabase, 69/100
9. Onlook, 69/100
10. Payload, 68/100
11. Sentry, 64/100
12. Cal.com, 63/100
13. Dub, 62/100

The same page invites users to run npx -y react-doctor@latest . on their own codebase and open a PR to add new entries, which makes the score feel less like a marketing screenshot and more like an ongoing public corpus.