clawpdf 🦞 — Pull the text and pixels out of PDFs

ClawPDF is an ESM-only PDFium WebAssembly library and CLI for Node.js and bundled browser apps. It extracts text, renders pages, and produces PNG bytes without runtime dependencies, native addons, postinstall scripts, or a canvas package.

Install

npm install clawpdf

Node.js 22 or newer is required. The package includes the clawpdf command and its PDFium WebAssembly runtime.

Quick start

Extract text from a PDF without writing a script:

$ npx clawpdf https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf
Dummy PDF file

Or open a local file from JavaScript and render its first page:

import { writeFile } from "node:fs/promises";
import { openPdf } from "clawpdf";

await using pdf = await openPdf("report.pdf");
console.log(pdf.text({ maxPages: 5 }));

const png = await pdf.page(1).png({ dpi: 144, forms: true });
await writeFile("page-1.png", png);

Page numbers in the CLI and library are one-based.

CLI

The CLI reads file paths, URLs, and standard input. Text goes to stdout; diagnostics go to stderr.

| Task | Command | | --- | --- | | Extract text | clawpdf report.pdf | | Read standard input | cat report.pdf \| clawpdf - | | Emit structured output | clawpdf report.pdf --json | | Render one page | clawpdf render report.pdf --page 1 -o page.png |

HTTP(S) downloads have no size cap by default. Opt into a byte budget with --fetch-max-bytes 10000000 or the library's fetchMaxBytes option. 0 disables the budget; the existing 30-second deadline still applies.

See the CLI reference for extraction modes, page selection, passwords, image budgets, inline terminal images, JSON output, and exit codes.

Text-first extraction

extractPdf() can return text, PNG fallback images, or both. In auto mode it always extracts text, then renders selected pages only when the text is shorter than minTextChars.

import { extractPdf } from "clawpdf";

const result = await extractPdf("report.pdf", {
  mode: "auto",
  maxPages: 20,
  minTextChars: 200,
  image: { dpi: 96, maxPixels: 4_000_000, forms: true },
});

console.log(result.text, result.images);

Image results contain raw PNG bytes. The optional clawpdf/adapters export converts them to data URLs or model-message content blocks. See extraction fallback for the modes, limits, and result shape.

Reuse an engine

openPdf() owns a private engine and suits one-off work. Servers should create one engine, reuse it across documents, and dispose it during shutdown:

import { createEngine } from "clawpdf";

export async function inspect(pdfBytes: Uint8Array) {
  await using engine = await createEngine();
  await using pdf = await engine.open(pdfBytes);

  console.log(pdf.metadata.title);
  console.log(pdf.page(1).text());
}

The Node entry point accepts byte arrays, file paths, URLs, and blobs. Bundled browser code should import the pre-wired clawpdf/browser entry point; see browser and bundler setup.

Documentation

• Loading PDFs covers inputs, document lifetimes, and passwords.
• Text extraction, page rendering, and PNG output cover the core document APIs.
• Password-protected PDFs covers the library and CLI flows.
• API reference lists exports, types, options, and typed errors.
• Package shape records the published files and dependency contract.
• PDFium provenance records the vendored release, checksum, notices, and refresh workflow.
• Performance contains the current local benchmark snapshot.

The full documentation site is available at clawpdf.dev.

Development

pnpm install
pnpm build
pnpm typecheck
pnpm test
pnpm test:package
pnpm docs:site

CI runs the same checks on Node.js 22, 24, and 26.

License

MIT. PDFium carries upstream BSD-style and Apache-2.0 notices; see THIRD_PARTY_NOTICES.md.