What Is React Doctor and What Does It Do for React Codebases?
Updated
React Doctor is an open-source tool from Million that helps developers find issues in their React codebases. It is one of the developer tools Million builds as part of its stated mission to "fix the web," alongside React Scan and ReactBench. If you maintain a React application and want a way to surface problems in your code, React Doctor is aimed at you. The specific checks it runs, its installation steps, and its output format are not detailed in the available source material, so treat the sections below as a map of what is confirmed versus what you will need to verify on the project itself.
What React Doctor is
Million describes React Doctor as an open-source tool that "has helped developers find issues in their codebases." Two things follow from that description:
- It is a diagnostic tool, not a framework or library. You run it against existing code rather than adopting it as a dependency of your app's runtime.
- It is open source. The code is publicly available, which means you can inspect what it checks and how, rather than trusting a black box.
Million's own framing places React Doctor in a family of tools: React Doctor and React Scan are named together as the open-source tools the company ships, while ReactBench, custom datasets, reinforcement learning environments, and traces are described separately as resources for training and evaluating coding agents.
The problem it addresses
React codebases accumulate issues that are easy to write and hard to notice: components that re-render more than they need to, patterns that work today but degrade as the app grows, and small mistakes that only show up under real usage. A tool that scans a codebase and reports problems gives you a starting point for cleanup that does not depend on someone remembering to look.
The source material does not enumerate the specific categories of issues React Doctor detects. What it does confirm is the outcome: developers have used it to find issues in their codebases. If you need the exact rule set before adopting it, check the project's own documentation rather than assuming coverage.
How it relates to React Scan and ReactBench
These three names appear together in Million's materials but serve different roles, and conflating them leads to wrong expectations.
| Tool | What it is for | Who it is for |
|---|---|---|
| React Doctor | Finding issues in a React codebase | Developers maintaining React apps |
| React Scan | A separate open-source tool from the same family | React developers |
| ReactBench | A benchmark built alongside datasets and RL environments to train and evaluate frontier coding agents | Teams working on coding agents |
The practical takeaway: React Doctor and React Scan are tools you use on your own code, while ReactBench is infrastructure for evaluating AI coding agents on realistic web development work. They come from the same company and share a mission, but they are not interchangeable and you would not adopt one as a substitute for another.
Who benefits most
React Doctor fits developers who already have a React codebase and want to find problems in it. That includes:
- Teams inheriting or auditing an existing React app who need a fast way to surface issues rather than reading every file.
- Developers preparing for a refactor who want a list of concrete problems to work from.
- Anyone evaluating Million's tooling who wants to start with the diagnostic tool before looking at the agent-training side of the company's work.
It is less relevant if you are not working in React, or if you are looking for a runtime performance monitor rather than a codebase scanner — that is closer to what React Scan is described as.
How to get started
The source material does not include installation commands, CLI flags, or configuration options, so the honest starting point is the project itself:
- Find the repository. React Doctor is open source and published by Million; the project page or Million's site is where the current instructions live.
- Read the README before running anything. Because the tool's checks are not documented in the material available here, the README is where you confirm what it looks for and what it expects as input.
- Run it against a branch, not your main working tree, the first time. Any codebase scanner can produce a long report; keeping the first run isolated makes it easy to discard noise.
- Triage the output by severity or frequency rather than fixing everything at once. The value is in the list, not in clearing it in one pass.
What is not confirmed here
To avoid overstating what is known: the available material does not specify React Doctor's installation method, supported React versions, rule categories, output format, license terms, or pricing. Million's site shows no pricing signals for its tools, but that is an absence of information, not evidence that everything is free or requires no account. Check the project's own pages for those details before planning around it.