DeepRead is an AI-native OCR platform that turns documents into high-accuracy structured data using multi-model consensus and automated human-review flagging.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install clawdhub-skill
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 clawdhub-skill 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?
DeepRead provides a production-grade document processing API that bridges the gap between raw documents and structured data. By leveraging a multi-model consensus approach, it achieves over 95% accuracy and automatically identifies uncertain fields for human review, reducing manual workloads by up to 90%. This platform is designed for developers who need reliable text extraction without the complexity of prompt engineering.
Integrating this tool within your Openclaw Skills ecosystem allows for seamless document digitization. It handles the heavy lifting of document rotation, multi-pass validation, and confidence scoring, ensuring that your automation pipelines receive clean, validated data every time.
DeepRead Use Cases
- Automating invoice processing to extract vendors, totals, and line items.
- Parsing retail receipts for merchant names, transaction dates, and itemized totals.
- Analyzing legal contracts to identify parties, effective dates, and specific terms.
- Digitizing paper forms into structured JSON for database ingestion.
- Building quality-critical applications where identifying uncertain extractions via the hil_flag is mandatory.
How DeepRead Works
- Document Submission: A PDF or image is uploaded via the API with an optional JSON schema defining the target fields.
- Automated Pre-processing: The system corrects document rotation and orientation for optimal reading.
- Multi-Pass Analysis: The document undergoes multiple validation passes using cross-model consensus to ensure maximum accuracy.
- Structured Extraction: Data is mapped to the requested schema, and each field is assigned a confidence score.
- Review Flagging: The system sets a hil_flag to true for any fields that are blurry, handwritten, or ambiguous.
- Asynchronous Delivery: Processed results are delivered via webhooks or polling once the 2-5 minute processing window is complete.
DeepRead Setup
To integrate this capability into your Openclaw Skills environment, start by obtaining an API key from the DeepRead dashboard.
Set your environment variable:
export DEEPREAD_API_KEY="sk_live_your_key_here"
For persistent configuration in Clawdbot, add the following to your clawdbot.config.json5:
{
"skills": {
"entries": {
"deepread": {
"enabled": true,
"apiKey": "sk_live_your_key_here"
}
}
}
}
DeepRead Data Schema & Taxonomy
The skill returns a comprehensive JSON object containing both raw and structured data. Key elements include:
| Component | Description |
|---|---|
text |
The full document content extracted as clean Markdown. |
data |
An object containing extracted fields (e.g., vendor, total) with associated confidence flags. |
hil_flag |
A boolean value indicating if a human should review the specific field. |
metadata |
Processing statistics including total fields, pages, and the percentage of fields requiring review. |
DeepRead Advanced Features
- Blueprints: Optimized, reusable schemas for specific document types that improve accuracy by 20-30%.
- Production Webhooks: Real-time notifications upon processing completion to eliminate polling overhead.
- Public Preview URLs: Generate unauthenticated links to share OCR results and previews for quick verification.
- Multi-Model Validation: Cross-references results across different AI models to ensure data integrity.
- Multimodal Quality Detection: Identifies low-quality, blurry, or obscured text using advanced vision analysis rather than simple rules.
SKILL.md
Loading
Related Openclaw Skills
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.
InstantDB Integration
A powerful Node.js skill for Openclaw that enables real-time database administration and live data synchronization using InstantDB.
DeepRead OCR & Structured Data Extraction
DeepRead is a production-grade OCR API that extracts structured JSON data from PDFs with AI-powered confidence scoring to minimize human review.
Audio Content Generator
A sophisticated workflow that transforms scripts into high-quality audiobooks, podcasts, or educational content using AI narration and ElevenLabs synthesis.
Featured*
