oxc (OXC Toolchain)

Experimental Rust-based JavaScript/TypeScript toolchain with linter, formatter, resolver, and transformer. Targets performance-critical CI/CD pipelines.

Status: Experimental (2026). API and rules may change between versions.

Dependencies:

{
  "oxlint": "latest"
}

Core Patterns

✅ REQUIRED: Understand Experimental Status

// ✅ CORRECT: Use oxc when you accept:
// - Rules may change between versions
// - Fewer rules than ESLint (~300 vs 1000+)
// - Limited plugin ecosystem
// - Breaking changes possible

// ❌ WRONG: Use as drop-in ESLint replacement
// Not all ESLint rules have oxc equivalents
// Plugin ecosystem is much smaller

✅ REQUIRED: Basic Setup

# Install
npm install -D oxlint

# Run linter
npx oxlint .

# With configuration
npx oxlint --config oxlintrc.json .

// oxlintrc.json
{
  "rules": {
    "no-unused-vars": "error",
    "no-explicit-any": "error",
    "no-var": "error",
    "prefer-const": "error",
    "eqeqeq": "error"
  }
}

✅ REQUIRED: Package.json Scripts

{
  "scripts": {
    "lint": "oxlint .",
    "lint:fix": "oxlint --fix ."
  }
}

When to Use oxc

Good fit:

• CI/CD where lint speed is critical (10x-100x faster than ESLint)
• Large monorepos where ESLint is slow
• As complement to ESLint (run oxc first for fast feedback, ESLint for full check)
• Projects willing to accept experimental tooling

Bad fit:

• Need specific ESLint plugins (react-hooks, import, etc.)
• Need stable, production-proven linting rules
• Team unfamiliar with experimental tools
• Need formatting (oxc formatter is early stage)

Performance Comparison

Approximate benchmarks. Actual performance varies by rule set and file content.

Available Rule Categories

{
  "rules": {
    // Correctness
    "no-unused-vars": "error",
    "no-undef": "error",
    "no-const-assign": "error",

    // Style
    "no-var": "error",
    "prefer-const": "error",

    // Suspicious
    "no-explicit-any": "error",
    "eqeqeq": "error",
    "no-debugger": "error",

    // TypeScript (built-in)
    "no-explicit-any": "error",
    "@typescript-eslint/no-unused-vars": "error"
  }
}

Integration Strategies

As ESLint Complement

// package.json — Run oxc first (fast), then ESLint (thorough)
{
  "scripts": {
    "lint:fast": "oxlint .",
    "lint:full": "eslint . --ext .ts,.tsx",
    "lint": "npm run lint:fast && npm run lint:full"
  }
}

As Standalone

{
  "scripts": {
    "lint": "oxlint .",
    "lint:fix": "oxlint --fix .",
    "format": "biome format --write ."
  }
}

oxc formatter is early stage. Pair with Biome or Prettier for formatting.

CI/CD Integration

# GitHub Actions — fast lint check
- name: Fast lint
  run: npx oxlint .

# Or as pre-check before full lint
- name: Quick lint (oxc)
  run: npx oxlint .
- name: Full lint (ESLint)
  run: npm run lint

Edge Cases

Missing rules: oxc has ~300 rules vs ESLint’s 1000+. Check available rules before migrating.

No plugin system: oxc doesn’t support ESLint plugins. React-specific and import-specific rules are built-in but limited.

Formatter status: oxc formatter is very early stage. Use Biome or Prettier for formatting.

Breaking changes: Pin versions in CI.

Comparison with Biome

Recommendation: For most projects, Biome is the better modern choice. Use oxc only when raw speed is the top priority and you accept the experimental trade-offs.

Related Topics

• See biome.md for stable modern alternative
• See eslint.md for mature linting
• See main code-quality/SKILL.md for decision tree

References

• oxc Documentation
• oxlint GitHub
• oxc Playground