Document Intelligence, Redefined

Parse Any Document
With Surgical Precision

Parsy goes beyond simple text extraction. It understands structure, preserves tables, extracts metadata, detects language, and delivers pixel-perfect markdown — all in your browser with zero data leaving your machine.

Start Parsing Free See Capabilities
ENGINE_VERSIONv3.4.1-local
WORKER_POOLMultiprocess / WebWorker
SANDBOX_ISOLATIONActive (Localhost)
TELEMETRY0.0% (Opt-out default)
parsy_output.md
// 01 / core_capabilities

Everything LlamaParse does,
done better

Native Layout Reconstruction

Faithfully recreates multi-column layouts, nested sections, headers and footers. Reading order is intelligently inferred even in complex documents.

Perfect Table Ingestion

Merges cells, detects headers, handles rowspan/colspan, and outputs clean Markdown tables, JSON arrays, or CSV files.

5 Configurable Formats

Export directly to Markdown, Plain Text, JSON, HTML, and CSV. Each format is strictly schema-validated for LLM downstream applications.

100% Private Processing

All processing is client-side or worker isolated. No files are uploaded to third-party endpoints. Zero risk of data exfiltration.

document_restruct_preview.cpp ● idle
Input Wireframe (PDF)
Output Structure (MD)
# Q4 Executive Summary
Revenue grew by 24% YoY.
## Performance Table
| Q3 | Q4 | Change |
|---|---|---|
| 114M | 142M | +24% |
// 02 / specs_matrix

Parse anything

FORMAT
PARSER LAYER
FEATURES EXTRACTED
ACCURACY
SANDBOXING
PDF PDF Document
Level 2a / 2b (OCR)
Text layouts, nested tables, metadata, OCR fallback
99.8%
Process isolated
DOCX Word Document
Level 2a (Mammoth AST)
Semantic styles, tables, list nesting
99.5%
Process isolated
CSV Excel/CSV
Level 2a (Delimited)
Tabular arrays, numeric parsing, column mappings
100%
In-memory
JSON JSON Data
Level 2a (AST Parser)
Validation, schema detection, clean formatting
100%
In-memory
HTML HTML Pages
Level 2a (DOM Stripper)
Boilerplate removal, article main-body extraction
98.7%
In-memory
XML XML Schemas
Level 2a (XML Parser)
Node hierarchies, namespaces, structure validation
99.1%
In-memory
// 03 / orchestration_flow

3-Level Orchestration Engine

Every document takes the fastest, most accurate path automatically.

🔍
Level 1 · Router
Byte inspection · text density · image ratio · rotation detection
→
⚡
Level 2a · Fast Path
Digital PDF · DOCX · HTML
Parallel page chunking
👁
Level 2b · Vision + OCR
Scanned pages · rotated · images
Tesseract LSTM parallel
→
⚙️
Level 3 · Normalizer
Heading tree · date ISO · table padding · schema validation
// 04 / parser_workspace

Drop your document.
Get perfect output.

Drop files here

or

PDF · DOCX · TXT · MD · CSV · JSON · HTML · XML
Processed locally — nothing is uploaded