Pydantic

Pydantic is the most widely used data validation library for Python, now evolved into an end-to-end AI engineering stack. Founded by Samuel Colvin, the library uses Rust-based core (pydantic-core, built with PyO3) for validation performance up to 17x faster than pure Python alternatives. Pydantic Validation provides type-safe data parsing using Python type hints, supporting complex nested models, custom validators, constrained types (conint, conlist, confloat with bounds), discriminated unions, recursive models, JSON Schema generation, and serialization with mode-specific field exclusion. As of 2026, the Pydantic stack includes four products: Pydantic Validation (the core validation library, over 20,000 stars), Pydantic AI (a framework for building production-grade AI agents with type-safe tool calling, structured outputs via function tools, dependency injection, graph-based workflows, and MCP protocol support), Pydantic Logfire (an observability platform based on OpenTelemetry for logs, traces, and metrics with LLM call tracing), and Pydantic Evals (code-first evaluation framework with assertions and datasets). Pydantic is trusted by organizations including Microsoft, Meta, NVIDIA, Cisco, JPMorgan Chase, NATO, Walmart, Duolingo, Atlassian, and Jane Street. The library powers major frameworks: FastAPI (API validation), LangChain (structured LLM outputs), OpenAI SDK (tool call parsing), and Amazon Bedrock.