DeepRead is a production-grade OCR API that extracts structured JSON data from PDFs with AI-powered confidence scoring to minimize human review.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install deepread
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install deepread using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is DeepRead OCR & Structured Data Extraction?
DeepRead is an advanced document processing skill designed to handle complex OCR tasks where accuracy is paramount. Unlike traditional OCR tools that fail silently, DeepRead utilizes a multi-pass pipeline and multi-model consensus to evaluate extraction quality. It effectively reduces manual data entry and human review by approximately 90% by highlighting only the specific fields that require verification.
By integrating this tool with other Openclaw Skills, developers can build robust automation pipelines for invoices, contracts, and forms. The system handles document rotation, orientation correction, and field-level confidence scoring, ensuring that your AI agents work with the most reliable data possible.
DeepRead OCR & Structured Data Extraction Use Cases
- Automated invoice processing to extract vendors, totals, and line items directly into accounting software.
- Parsing receipts and merchant data for expense management systems using Openclaw Skills.
- Analyzing legal contracts to identify parties, dates, and specific termination clauses automatically.
- Digitizing high volumes of physical forms into structured JSON datasets for database entry.
- Building quality-critical document workflows where human-in-the-loop review is only triggered by AI uncertainty.
How DeepRead OCR & Structured Data Extraction Works
- Upload a document (PDF, JPG, or PNG) via the DeepRead API endpoint.
- The pipeline performs automated rotation correction and multi-pass OCR processing.
- Multimodal AI models cross-validate the extracted text against the original document image.
- Structured data is generated based on a provided JSON schema or a specialized Blueprint.
- The system assigns a hil_flag (Human-in-the-Loop) to any fields where extraction confidence is low.
- Results are delivered via a webhook notification or retrieved through an asynchronous polling request.
DeepRead OCR & Structured Data Extraction Setup
1. Obtain API Key
Sign up at the DeepRead Dashboard and export your key to your environment:
export DEEPREAD_API_KEY="sk_live_your_key_here"
2. Configure Clawdbot
Add the skill to your Openclaw Skills configuration in clawdbot.config.json5:
{
skills: {
entries: {
"deepread": {
enabled: true,
apiKey: "sk_live_your_key_here"
}
}
}
}
3. Process a Document
Use cURL to upload a file and define your extraction schema:
curl -X POST https://api.deepread.tech/v1/process \
-H "X-API-Key: $DEEPREAD_API_KEY" \
-F "[email protected]" \
-F 'schema={"type":"object","properties":{"total":{"type":"number"}}}'
DeepRead OCR & Structured Data Extraction Data Schema & Taxonomy
DeepRead organizes data into a structured JSON response containing text, extracted fields, and quality metadata. This structure is ideal for developers using Openclaw Skills to automate data entry.
| Component | Type | Description |
|---|---|---|
text |
String | The full document content extracted as clean Markdown format. |
data |
Object | Key-value pairs extracted based on your custom JSON schema. |
hil_flag |
Boolean | True if the AI suggests human review for that specific field or page. |
pages |
Array | Page-by-page breakdown including individual text blocks and quality flags. |
metadata |
Object | Processing statistics including confidence scores and review percentages. |
DeepRead OCR & Structured Data Extraction Advanced Features
- Blueprints: Create reusable, optimized schemas that improve extraction accuracy by up to 30% for specific document types using ground-truth training data.
- Webhooks: Enable real-time notifications for production-grade Openclaw Skills workflows to receive data immediately upon processing completion.
- Public Previews: Generate shareable, unauthenticated URLs to view OCR results and images in a web-friendly previewer.
- Multi-Model Consensus: Leverage cross-validation between different AI models to ensure the highest possible reliability for critical data points.
- Page-Level Breakdown: Access detailed OCR results for each individual page, including specific reasons for quality flags.
SKILL.md
Loading
Related Openclaw Skills
DeepRead
DeepRead is an AI-native OCR platform that turns documents into high-accuracy structured data using multi-model consensus and automated human-review flagging.
UCTOO API Integration Skill
An advanced Openclaw Skills integration that translates natural language into UCTOO backend API requests with automated token management.
UCM: API Marketplace for AI Agents
A comprehensive API marketplace that empowers AI agents to perform external tasks like web searching, image generation, and code execution through a single unified interface.
APort Agent Guardrail
A deterministic pre-action authorization engine that verifies AI agent tool calls against a security policy before execution.
Audio Content Generator
A sophisticated workflow that transforms scripts into high-quality audiobooks, podcasts, or educational content using AI narration and ElevenLabs synthesis.
Clawdbot Sync
A robust synchronization utility for maintaining consistent memory, user profiles, and skills across multiple distributed AI agent instances.
METADATA
Featured*
