Pyrefly is a high-performance Python type checker and language server developed by Meta. Designed as the successor to Pyre, it combines extreme speed with robust IDE features to provide developers with instant feedback on type-related errors. It helps maintain large Python codebases by catching bugs early, improving code quality, and serving as living documentation that is automatically validated by the static analysis engine. Pyrefly is engineered to handle massive projects, demonstrating the ability to type-check millions of lines of code per second.

Functionality includes both CLI-based type checking and deep IDE integration, facilitating seamless workflows for developers in environments like VS Code, PyCharm, Cursor, Neovim, and others. It features advanced type inference, structural typing support, and sophisticated variance inference, allowing it to understand complex Python code structures effectively. The tool also provides automation utilities for generating stub files, measuring type coverage, and automating type annotations.

Some of the key features are:

• Extreme Performance: Optimized to type-check over 1.85 million lines of code per second.
• Advanced Inference: Infers types from code literals and function signatures with high precision.
• IDE Integration: Provides lightning-fast autocomplete, go-to-definition, and inline documentation via LSP.
• Automated Migration: Includes built-in support to migrate existing configurations from Mypy and Pyright.
• Customizable Presets: Offers various strictness levels including basic, strict, and legacy modes.
• Agentic Loop Support: Designed to integrate easily into CI/CD pipelines and AI-driven development workflows.
• Structural Typing: Full support for Protocols, TypedDicts, and dataclasses to match Python's dynamic nature.

Operationally, Pyrefly is installed via pip or common package managers and is initialized with a command-line tool that detects existing configurations. Once active, it monitors the project for changes, providing real-time diagnostics as code is written. It can be enforced in continuous integration environments using official GitHub Actions, ensuring that no type-unsafe code is merged into production branches. The tool's configuration system allows for fine-grained control over error reporting, path includes/excludes, and environmental settings.

Some common use cases include:

• Large Scale Refactoring: Safely making wide-reaching changes across thousands of files with immediate type safety validation.
• AI-Assisted Development: Catching type bugs in code generated by AI coding agents before execution.
• Project Migration: Converting legacy Python projects from Mypy or Pyright to Pyrefly while preserving existing settings.
• CI/CD Enforcement: Automating type checking as a gate in pull request reviews to maintain consistent code quality standards.
• Library Maintenance: Improving type completeness in popular open-source libraries to benefit the wider Python ecosystem.