Skip to main content

Quickstart

Get started with Docsray in 5 minutes! This guide will walk you through basic usage and show you how to extract everything from documents.

Important: Docsray is an MCP server that works through Claude Desktop or other MCP clients. The commands shown below are what you type in Claude, not in your terminal. See How It Works for details.

Your First Document Analysis​

1. Quick Overview (Peek)​

Start by getting a quick overview of any document in Claude:

# Local file
Peek at ./invoice.pdf

# URL
Peek at https://example.com/document.pdf

Response shows:

  • Page count and file size
  • Available extraction formats
  • Provider capabilities
  • Document metadata

2. Extract Everything (Xray)​

Use the power of AI to extract ALL data:

Xray invoice.pdf with provider llama-parse

This returns:

  • All text content
  • All tables structured
  • All images with descriptions
  • All entities recognized
  • Complete document hierarchy

3. Get Specific Content (Extract)​

Extract content in your preferred format:

# As markdown
Extract pages 1-5 from report.pdf as markdown

# As JSON with tables
Extract all tables from financial.pdf as JSON

# Specific pages
Extract page 3 from contract.pdf

Maximum Extraction Example​

Here's how to get EVERYTHING from a document in Claude:

Xray document.pdf with provider llama-parse and custom instructions: 
'Extract ALL possible information including: 
1) Complete text content preserving exact formatting
2) All tables with complete data
3) All images with descriptions
4) Complete document metadata
5) Full document structure
6) All entities (people, orgs, dates, amounts)
7) Page-by-page layout information'

Common Use Cases​

# Extract all parties and terms
Xray contract.pdf and extract all parties, dates, and obligations

# Find specific clauses
Seek to 'termination clause' in agreement.pdf

Financial Reports​

# Extract financial metrics
Xray 10-k.pdf for revenue, growth rates, and risk factors

# Get all tables
Extract all tables from earnings.pdf as JSON

Academic Papers​

# Map document structure
Map the structure of research-paper.pdf

# Extract methodology
Extract the methodology section from paper.pdf

Invoice Processing​

# Extract key data
Xray invoice.pdf and extract vendor, amount, date, and line items

# Get as structured data
Extract invoice.pdf as JSON with tables

Working with Providers​

Auto-Selection (Default)​

Let Docsray choose the best provider:

Analyze document.pdf  # Automatically selects provider

LlamaParse (Comprehensive)​

For deep analysis and AI-powered extraction:

Xray document.pdf with provider llama-parse

Features:

  • Entity recognition
  • Custom instructions
  • Table structure preservation
  • Image extraction
  • 5-30 second processing

PyMuPDF (Fast)​

For quick text extraction:

Extract document.pdf with provider pymupdf4llm

Features:

  • Sub-second processing
  • Basic markdown
  • Text extraction
  • No API required
Seek to page 10 in manual.pdf

Find Sections​

Seek to 'Introduction' section in thesis.pdf

Search Content​

Search for 'payment terms' in contract.pdf

Tips for Best Results​

1. Start with Peek​

Always peek first to understand the document:

Peek at document.pdf

2. Use Specific Instructions​

Be specific about what you want:

Extract all email addresses and phone numbers from contact.pdf

3. Leverage Caching​

Results are cached - subsequent requests are instant:

Xray report.pdf  # First time: 10 seconds
Xray report.pdf  # Second time: instant

4. Handle Large Documents​

Process specific pages for large documents:

Extract pages 1-10 from large-manual.pdf

Interactive Example​

Try this complete workflow in Claude:

# 1. Check what's in the document
Peek at sample.pdf

# 2. Map the structure
Map the structure of sample.pdf

# 3. Extract everything with AI
Xray sample.pdf with provider llama-parse

# 4. Get specific content
Extract tables from sample.pdf as JSON

# 5. Navigate to specific sections
Seek to page 5 in sample.pdf

Next Steps​

Ready to extract everything from your documents? Start using Docsray with Claude now!