From parsing and rendering to XFA forms, compliance, OCR, and digital signatures — in one SDK, across every language.
Compile to a ~3 MB Brotli-compressed WASM bundle (~10 MB uncompressed) that runs PDF processing entirely in the browser. No server uploads, complete privacy.
PDF/A archiving, PDF/UA accessibility, and XRechnung/ZUGFeRD e-invoicing support out of the box.
Create and verify PAdES signatures with LTV support, timestamping, and full certificate chain validation.
Local OCR with ocrs or cloud adapters for Mistral, Google, AWS, and Azure. Extract text from scanned documents. OCR is in active development — contact us for production use cases.
Pure Rust — no JVM, no C++ runtime, no system library dependencies. Ship a single binary. Deploy to containers, serverless functions, or air-gapped servers without compatibility headaches.
The SDK runs in production on untrusted, arbitrary PDFs from end users. That's a large attack surface. Rust eliminates entire classes of vulnerability at compile time.
Rust's type system and borrow checker make buffer overflows and use-after-free impossible. Memory safety is enforced at compile time — not by a GC, not at runtime.
PDFium, Ghostscript, and libpoppler accumulate dozens of CVEs per year. PDFluent has no C or C++ dependencies — there is no native layer to exploit.
Pure Rust means no JNI, no FFI, no dynamic linking to system libs. The binary runs inside seccomp or a WASM sandbox without any special configuration.
Memory-unsafe runtimes (C++, JVM) introduce classes of vulnerabilities — buffer overflows, use-after-free, deserialization gadgets — that Rust eliminates by design. None of these attack vectors exist in PDFluent.
Dynamically-linked native PDF libraries load shared objects (.so/.dll) into the host process; a vulnerability in the library compromises the process. PDFluent ships as a single statically-linked binary with no runtime native dependencies.
The entire codebase is Rust — no native blobs, no minified JS, no binary-only components. Security auditors can read every line that runs.
* Performance claims measured on AMD EPYC 7502P, single-threaded, 20,000-PDF corpus. Full methodology →
30-day evaluation — full SDK, no watermarks, no credit card.