reducto-cli

Reducto Claude Code Plugins

Parse, extract, and edit PDFs, images, spreadsheets, and Office documents directly inside Claude Code — powered by Reducto's document intelligence API.

Reducto turns unstructured documents into structured, LLM-ready data using agentic OCR and vision-language models. This plugin brings those capabilities into your Claude Code workflow so you can process invoices, contracts, medical records, financial filings, and more without leaving your terminal.

Installation

Via Vercel Skills CLI (recommended)

npx skills add reductoai/claude-plugins

Via Claude Code Plugin System

Add this marketplace to Claude Code:

/plugin marketplace add reductoai/claude-plugins

Then install the plugin:

/plugin install reducto-cli@reducto-plugins

Quick Start

# 1. Authenticate with your Reducto account
uvx --from reducto-cli reducto login

# 2. Parse a PDF into clean Markdown
uvx --from reducto-cli reducto parse document.pdf

# 3. Extract structured JSON from an invoice
uvx --from reducto-cli reducto extract invoice.pdf --schema schema.json

# 4. Fill out a form with natural language
uvx --from reducto-cli reducto edit form.pdf --instructions "Fill in name as 'John Doe'"

Plugin: reducto-cli

Parse — Convert Documents to Markdown

Transform any supported document into clean, structured Markdown with optional YAML metadata. Parsed output preserves document layout, tables, figures, and semantic structure.

uvx --from reducto-cli reducto parse document.pdf
uvx --from reducto-cli reducto parse ./documents/   # batch process a directory

Options:

Flag	Description
--agentic	Maximum accuracy mode — uses agentic OCR for tables, text, and figures (slower)
--change-tracking	Outputs <s>, <u>, and <change> tags for tracked revisions
--highlights	Includes highlighted text in output
--hyperlinks	Preserves embedded hyperlinks
--comments	Includes document comments and annotations

# High-accuracy parse with full metadata
uvx --from reducto-cli reducto parse contract.pdf --agentic --change-tracking --comments

Output files are written as <filename>.parse.md.

Extract — Pull Structured Data with JSON Schema

Define the data you need using a JSON Schema and Reducto extracts it into typed JSON. The top-level schema type must be object.

uvx --from reducto-cli reducto extract invoice.pdf --schema schema.json

Example schema for invoice extraction:

{
  "type": "object",
  "properties": {
    "invoice_number": { "type": "string" },
    "date": { "type": "string" },
    "vendor": {
      "type": "object",
      "properties": {
        "name": { "type": "string" },
        "address": { "type": "string" }
      }
    },
    "line_items": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "description": { "type": "string" },
          "quantity": { "type": "number" },
          "unit_price": { "type": "number" },
          "total": { "type": "number" }
        },
        "required": ["description", "quantity", "unit_price", "total"]
      }
    },
    "total": { "type": "number" }
  },
  "required": ["invoice_number", "line_items", "total"]
}

You can also pass schemas inline:

uvx --from reducto-cli reducto extract receipt.pdf --schema '{"type":"object","properties":{"total":{"type":"number"},"date":{"type":"string"}}}'

Output files are written as <filename>.extract.json.

Edit — Modify Documents with Natural Language

Programmatically fill forms, update fields, and modify documents using plain English instructions. Edited files are returned in their original format.

uvx --from reducto-cli reducto edit application.pdf -i "Fill out: Name: Jane Smith, Email: jane@example.com"
uvx --from reducto-cli reducto edit contract.pdf -i "Set the effective date to March 1, 2025 and the client name to Acme Corp"
uvx --from reducto-cli reducto edit ./forms/ -i "Check the 'Approved' box and sign with today's date"

Output files are written as <filename>.edited.<ext>.

Supported File Types

Category	Formats
PDF	.pdf
Images	.png, .jpg, .jpeg
Word Documents	.doc, .docx
Presentations	.ppt, .pptx
Spreadsheets	.xls, .xlsx

Reducto handles scanned documents, faxes, and handwritten content through its OCR engine, and supports 100+ languages including mixed-language documents.

Use Cases

Invoice and Receipt Processing

Parse a folder of invoices, extract line items, totals, and vendor info into JSON, then feed the structured data into your accounting pipeline.

uvx --from reducto-cli reducto extract ./invoices/ --schema invoice_schema.json