OpenSSL is a production-grade cryptographic library and toolkit that implements the TLS, DTLS, and QUIC protocols alongside a comprehensive general-purpose cryptographic library (libcrypto). The codebase is organized around a modular provider architecture: core cryptographic algorithms (ciphers, digests, key exchanges, signatures, etc.) are implemented as loadable providers (default, FIPS, legacy, base, null), with a unified EVP API that dispatches operations to the appropriate provider. The library exposes a vast public API (hundreds of functions in include/openssl/) for secure communications, certificate handling (X.509, PKCS#12, CMS, OCSP), and low-level crypto primitives. The command-line tool (apps/openssl.c) provides administrative and diagnostic commands for key generation, certificate management, encryption, and protocol testing. The architecture is layered: the BIO abstraction handles I/O (files, sockets, memory buffers), the SSL/TLS/QUIC state machines manage handshakes and record layer processing, and the provider framework isolates algorithm implementations from the core library. Key entry points include library initialization (crypto/init.c), provider loading (crypto/provider_core.c), SSL context creation (ssl/ssl_lib.c), and the EVP interface (crypto/evp/). Data flows typically start with an application calling EVP or SSL functions, which route through the provider dispatch to platform-optimized implementations (e.g., AES-NI, ARM NEON, RISC-V vector extensions). The codebase includes extensive fuzzing harnesses (fuzz/), a comprehensive test suite (test/), and demos (demos/) illustrating common use cases. External dependencies are minimal: the core library relies only on OS threading (pthreads, Windows threads) and basic system calls. Optional dependencies include compression libraries (zlib, zstd, brotli) and external test frameworks (tlsfuzzer, NSS, BoringSSL for interop testing). The project actively maintains FIPS 140-3 validation modules and supports post-quantum algorithms (ML-DSA, ML-KEM, SLH-DSA, LMS) alongside legacy algorithms. Platform-specific hardware acceleration is handled through compile-time detection and runtime capability checks (e.g., crypto/armcap.c, crypto/riscvcap.c).