LaunchKit · 2026
Back to Skills

docstrange

Document extraction API by Nanonets.

21
2.9k downloads
by @shhdwi

Setup & Installation

openclaw skills install @shhdwi/docstrange

Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:

npx clawhub install docstrange

Version History

v1.0.2Feb 11, 2026 - Version note by skill developer:

- Added detailed recommendations to use environment variables for API key management, with security precautions for storing keys in configuration files. - Introduced a new Security & Privacy section covering data handling, privacy policy review, PII guidance, API key handling, compliance, and operational safeguards. - Included best practices for limiting permissions, rotating API keys, and monitoring usage. - Documented file size limits and advice on endpoint usage for large documents. - Updated and expanded instructions and notes for improved security and operational rigor. - No functional changes to the API or code itself; documentation and instructions only.

What This Skill Does

DocStrange is a document extraction API by Nanonets. It converts PDFs and images to markdown, JSON, or CSV, returning per-field confidence scores for each extracted value.

Per-field confidence scores (0-100) let you automatically flag uncertain extractions for manual review rather than trusting all output blindly.

When to use it

  • Extracting invoice fields from supplier PDFs
  • Parsing receipts for expense reports
  • Converting scanned bank statements to structured data
  • Digitizing paper forms into JSON
  • OCR-ing printed documents for search indexing

Example Workflow

Here's how your AI assistant might use this skill in practice.

INPUT

User asks: Extract the invoice number, date, vendor, and total from this invoice PDF

AGENT
  1. 1Receives the PDF file path from the user
  2. 2Sends a POST request to /api/v1/extract/sync with output_format=json and field list ["invoice_number", "date", "vendor", "total"]
  3. 3Adds include_metadata=confidence_score to the request
  4. 4Parses the response and flags any field with a confidence score below 80 for manual review
OUTPUT

JSON object with invoice_number, date, vendor, and total values, each annotated with a 0-100 confidence score

Requirements

Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.

Nanonets DocStrange accountDOCSTRANGE_API_KEY environment variable