Imported from mui/material-ui-docs (AGENTS.md). Install upstream with npx skills add mui/material-ui-docs. Copyright stays with the author.

AGENTS.md

This file provides guidance for AI agents working with code in this repository.

Package Manager

Only pnpm is supported (yarn/npm will fail). Use the -F flag for workspace operations:

pnpm -F @mui/material add some-package    # Add dependency to a package
pnpm -F @mui/material build               # Build a specific package

Never use cd to navigate into package directories for commands.

Common Commands

Development

pnpm install                  # Install deps if necessary
pnpm docs:dev                 # Start docs dev server only

Building

pnpm release:build            # Build all packages (except docs)
pnpm docs:build               # Build documentation site

Testing

pnpm test:unit                # Run all unit tests (jsdom)
pnpm test:unit ComponentName  # Run tests matching pattern
pnpm test:unit -t "test name" # Grep for specific test name
pnpm test:browser             # Run tests in real browsers (Chrome, Firefox, WebKit)
pnpm test:e2e                 # End-to-end tests
pnpm test:regressions         # Visual regression tests

Code Quality

pnpm prettier                 # Format staged changes
pnpm eslint                   # Lint with cache
pnpm typescript               # Type check all packages

API Documentation

After changing component props or TypeScript declarations:

pnpm proptypes && pnpm docs:api

Docs demos

Always author the TypeScript version of the demos. To generate the JavaScript variant, run:

pnpm docs:typescript:formatted

Architecture

This is a monorepo managed by Lerna with Nx for caching. Key packages:

• @mui/material - Core Material UI components
• @mui/system - Styling system (sx prop, styled, theme)
• @mui/lab - Experimental components (new components go here first)
• @mui/icons-material - Material Design icons
• @mui/utils - Internal utilities
• @mui/styled-engine - CSS-in-JS abstraction (Emotion by default)

Internal packages (not published): @mui-internal/*, @mui/internal-*

Code Conventions

TypeScript

• Use interface (not type) for component props
• Export {ComponentName}Props interface from component files
• Path aliases available: @mui/material → ./packages/mui-material/src

Errors

These guidelines only apply for errors thrown from public packages.

Every error message must:

1. Say what happened - Describe the problem clearly
2. Say why it's a problem - Explain the consequence
3. Point toward how to solve it - Give actionable guidance

Format:

• Prefix with MUI:
• Use string concatenation for readability
• Include a documentation link when applicable (https://mui.com/r/...)

Error Minifier

Use the /* minify-error */ comment to activate the babel plugin:

throw /* minify-error */ new Error(
  'MUI: Expected valid input target. ' +
    'Did you use a custom `inputComponent` and forget to forward refs? ' +
    'See https://mui.com/r/input-component-ref-interface for more info.',
);

The minifier works with both Error and TypeError constructors.

After Adding/Updating Errors

Run pnpm extract-error-codes to update docs/public/static/error-codes.json.

Important: If the update created a new error code, but the new and original message have the same number of arguments and semantics haven't changed, update the original error in error-codes.json instead of creating a new code.

Component Structure

packages/mui-material/src/Button/
├── Button.tsx           # Component implementation
├── Button.d.ts          # TypeScript declarations (for JSDoc API docs)
├── Button.test.js       # Unit tests
├── buttonClasses.ts     # CSS classes
└── index.ts             # Public exports

Testing

• Use createRenderer() from @mui/internal-test-utils
• Use Chai BDD-style assertions (expect(x).to.equal(y))
• Custom matchers: toErrorDev(), toWarnDev() for console assertions
• Prefer testing components with full interactions using user.* methods. Avoid fireEvent and setProps if possible.
• If tests require the browser because, for example, they require layout measurements, restrict it to the Chromium env by using it.skipIf(isJsdom()) or describe.skipIf(isJsdom()) (search other tests for example usage if unsure).

import { createRenderer } from '@mui/internal-test-utils';

describe('Button', () => {
  const { render } = createRenderer();

  it('renders children', async () => {
    const handleClick = vi.fn();
    const { getByRole, user } = render(<Button onClick={handleClick}>Hello</Button>);

    const button = getByRole('button');
    expect(button).to.have.text('Hello');

    await user.click(button);
    expect(handleClick).toHaveBeenCalledTimes(1);
  });
});

Accessibility Testing

axe-core runs inside the visual-regression Playwright loop (test/regressions/index.test.js) — no separate browser session. Screenshots and a11y are independent: a demo can opt out of one and still run the other.

Key files:

• test/regressions/demoMeta.ts — SCREENSHOT_RULES and A11Y_RULES arrays, matched last-wins (no inheritance: overrides restate every field) against docs/data/material/components/{slug}/{Demo} (minimatch globs).
• test/regressions/a11y/axe.ts — asserts color-contrast and link-in-text-block by default, or all exercised axe rules when the matching a11y rule sets assertions: 'all'; skipAssertions suppresses selected rule assertions.
• test/regressions/a11y/a11yReporter.ts — writes one file per slug at docs/data/material/components/{slug}/{slug}.a11y.json. Each file is keyed by demo name, then by axe rule ID. Each rule records a status (pass, fail, or incomplete) and WCAG tags.

Enroll a component (slug-wide, or narrow with brace-glob):

// test/regressions/demoMeta.ts
{ test: 'docs/data/material/components/alert/*', enabled: true, skipAssertions: ['color-contrast'] },
{ test: 'docs/data/material/components/buttons/{BasicButtons,ColorButtons}', enabled: true, assertions: 'all' },

Override a specific demo: append a per-demo rule after the slug-wide rule (last-match-wins; the override must restate every field it wants):

{ test: 'docs/data/material/components/popover/AnchorPlayground', enabled: false }, // Redux isolation

Run pnpm test:regressions to refresh the *.a11y.json files. CI fails if any are stale.

For local iteration, scope the run with vitest's -t test-name filter (matched against the it() strings, which contain the route). Non-matching tests are skipped — their bodies don't execute, so the browser never navigates to those routes.

# in one terminal
pnpm test:regressions:server

# in another — note no `--`, pnpm forwards args directly
pnpm test:regressions:run -t '/docs-components-buttons/'              # one slug
pnpm test:regressions:run -t '/docs-components-buttons/BasicButtons$' # one demo
pnpm test:regressions:run -t '/docs-components-(buttons|chips)/'      # multiple slugs

Filtered runs only refresh the matched slugs' *.a11y.json. Run the unfiltered pnpm test:regressions before pushing.

Imports

Use one-level deep imports to avoid bundling entire packages:

import Button from '@mui/material/Button'; // Good
import { Button } from '@mui/material'; // Avoid in packages

Agent Skills

Packaged guidance for common integration topics lives under skills/. Each skill is a self-contained directory:

Read the relevant AGENTS.md when helping users with those topics.

Pre-PR Checklist

1. pnpm prettier - Format code
2. pnpm eslint - Pass linting
3. pnpm typescript - Pass type checking
4. pnpm test:unit - Pass unit tests
5. If API changed: pnpm proptypes && pnpm docs:api
6. If demos changed: pnpm docs:typescript:formatted
7. If .md files changed: pnpm vale <file1> <file2> ... - Check prose style and grammar

PR Title Format

[component] Imperative description

Examples:

• [button] Add loading state
• [docs] Fix typo in Grid documentation