The AI-nativePDF platform.
One embeddable viewer that renders, annotates, fills, redacts, signs, and compares — with a built-in AI copilot that summarises, explains, extracts, and parses every document, citing the exact page it read.
A copilot that reads every page — and shows its work.
Five AI surfaces live inside the viewer: streaming summaries, selection-level explain and translate, typed field extraction, full structural parsing, and vision for scans. Every answer is grounded with page-level citations you can click.
S.01 — The whole document, distilled while it streams.01 / 05
Answers with receipts
Every AI response carries page citations — click one and the viewer jumps to the page and flashes the exact source rectangle. Solid, dotted, or dashed chips tell you whether the evidence came from the text layer, OCR, or vision.
Works on scans too
Image-only pages are rasterised and routed to the vision model automatically. Paper contracts, faxes, and photographed receipts get the same treatment as born-digital PDFs.
Your model, not ours
Anthropic, Gemini, any OpenAI-compatible gateway, or a fully on-prem MCP server — one config object. TruPDF never proxies your prompts and disables cleanly when no key is set.
Enterprise guardrails
Proxy mode keeps API keys out of the browser. A prompt-check hook scrubs PII before anything leaves, every call lands in the audit log, and results cache locally in IndexedDB.
Everything a document workflow needs. Modular, typed, self-owned.
Each capability ships as its own workspace package behind one typed SDK — use the full viewer or compose engines directly.
One dependency. One component. Shipping this afternoon.
First-class React, with Vue, Angular, and Svelte wrappers mounting the same viewer implementation — including the AI panel.
I.01 — Install & mount
pnpm add @trupdf/react
import { TruPdfViewer } from '@trupdf/react';
<TruPdfViewer
src="/contracts/master-services.pdf"
license={process.env.TRUPDF_LICENSE}
aiConfig={{ provider: 'anthropic', proxyUrl: '/api/ai' }}
/>I.02 — Framework wrappers
The PDF.js worker ships inlined — no asset copying, no CDN, no bundler gymnastics. TypeScript-strict definitions across the whole public API.
Leaving your old viewer shouldn't cost you a rewrite.
TruPDF was built as a drop-in replacement: your documents, annotations, and form data carry over untouched.
Zero-migration switch
Existing PDFs, XFDF annotations, form data, signed and encrypted documents keep working as-is. Drop-in replacement for the legacy commercial WebViewer — no client rework.
Cross-viewer fidelity
Annotations are written to the PDF 1.7 / ISO 32000-2 spec, so what you author in TruPDF renders identically in Acrobat, Chrome, Firefox, Edge, and Preview.
Sensitive ops, done right
Structural redaction that physically removes bytes, PKI signatures with chain validation, and audit logging before any result leaves the engine.
31 languages, RTL support
Native-locale labels with direction flips for Arabic, Hebrew, and Persian — plus custom translation hooks for everything else.
WCAG 2.2 AA accessibility
ARIA live regions, keyboard navigation, structural reading order in secure mode, focus traps, and skip-link helpers ship by default.
Self-owned engine
No banned-license dependencies, no per-seat metering surprises, manual licensing, and direct support from the team that wrote every line.
Next step
Put an AI copilot inside your PDF viewer this week.
Get a 15-day trial license in under a minute. No credit card, no sales call — and the AI panel works with whatever model provider you already use.