A powerful skill to extract structured, typed JSON data from PDF or scanned purchase orders, including line items and confidence flags.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install deepread-purchase-orders
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-purchase-orders 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 Purchase Orders?
The DeepRead Purchase Orders skill transforms complex, unstructured PDF and scanned purchase orders into clean, developer-friendly JSON. By extracting essential fields like purchase order numbers, line items, buyer/supplier details, delivery dates, and totals, it removes the manual data entry bottleneck from your business workflows. As part of the Openclaw Skills library, this tool enables your AI agents to programmatically parse and process transactional documents with high precision.
With features like per-field confidence flags and page-location tracking, this skill is designed for production integrations that require human-in-the-loop review only when necessary. It empowers developers to build automated procurement pipelines, spend analysis engines, and robust accounts payable systems that seamlessly integrate with existing ERP platforms.
DeepRead Purchase Orders Use Cases
- Procurement automation: Programmatically ingest purchase orders into ERP, CRM, or custom database platforms without manual data entry.
- Accounts Payable (AP) matching: Combine this with invoice data to run automated 2-way and 3-way matching workflows to flag discrepancies before processing payments.
- Supplier and vendor tracking: Monitor incoming ordered SKUs, unit pricing, and delivery commitments across multiple suppliers in real time.
- Enterprise spend analysis: Aggregate line-item data across thousands of historical or ongoing purchase orders to uncover cost-saving opportunities.
How DeepRead Purchase Orders Works
- Document Submission: The agent initiates a POST request sending the PDF or scanned purchase order document to the DeepRead API.
- AI-Powered Extraction: The platform analyzes the document, extracting the PO number, line items, totals, and metadata while assigning confidence ratings.
- Asynchronous Polling: The agent polls the api endpoint until the status transitions to completed.
- Structured JSON Output: A strictly typed JSON object is returned, identifying the location of each extracted field and highlighting fields requiring review.
- Integration & Reconciliation: The structured output is fed into downstream Openclaw Skills or local code for verification (e.g., comparing totals against a corresponding invoice).
DeepRead Purchase Orders Setup
To use this skill within the Openclaw Skills framework, you need to set up your API credentials:
- Sign up or log in to the DeepRead Dashboard to retrieve your API key.
- Configure your environment variable:
export DEEPREAD_API_KEY=\"sk_live_your_key_here\"
DeepRead Purchase Orders Data Schema & Taxonomy
JSON Output Schema
The extracted payload conforms to the following schema format:
| Field | Type | Description |
|---|---|---|
po_number |
string | The unique purchase order identifier. |
order_date |
string (YYYY-MM-DD) | The date the purchase order was issued. |
supplier_name |
string | Name of the supplier or vendor. |
buyer_name |
string | Name of the buying entity or ordering company. |
delivery_date |
string or null | Requested delivery date. |
ship_to |
string or null | Shipping destination address. |
currency |
string | Currency code (e.g., USD, EUR). |
total |
number | The absolute total order amount. |
line_items |
array | List of items containing SKU, description, quantity, unit price, and total line amount. |
Per-Field Metadata
Each extracted field includes a metadata wrapper containing:
needs_review: A boolean flag indicating if human verification is recommended based on API confidence.location: Specifies the exact page number where the key-value pair was discovered.
DeepRead Purchase Orders Advanced Features
- Automated 2-Way & 3-Way Matching: Seamlessly pair this with other Openclaw Skills like
deepread-invoiceto match purchase orders against invoices and receipts dynamically in Python. - Human-In-The-Loop (HITL) Routing: Utilize the built-in
needs_reviewflags to isolate low-confidence extractions and automatically queue them for manual review. - Custom Supplier Blueprints: Build structured templates or optimizers within the DeepRead Dashboard to handle highly non-standard or recurring supplier formats with higher accuracy.
- Generous Free Tier: Process up to 2,000 pages per month for free, making it ideal for scaling testing environments before moving to production.
SKILL.md
Loading
Related Openclaw Skills
HR Candidate Discovery and Screening
An AI-powered candidate sourcing, vetting, and outreach automation workflow powered by DeepSeek and Lark Mail, designed to keep human recruiters strictly in control.
Operation Gladio
An AI-powered geopolitical research toolkit designed to trace historical alliances, deep-state finance, and the legacy of Operation Gladio stay-behind networks.
Earn Hunter
Earn Hunter is an automated yield monitor for OKX Flash Earn, Fixed Earn, and Flexible Earn opportunities that triggers notifications across Telegram, Lark, or your terminal session.
Into Thin Air Narrative & Decision-Making Toolkit
An AI-powered narrative toolkit that maps Jon Krakauer's Into Thin Air to high-stakes decision-making, mountaineering history, and survival ethics.
Market Entity Statistics API
Access real-time enterprise statistics, business registrations, and regional growth metrics for intelligent business analyses.
AutoPostOnline Agents
AutoPostOnline Agents is an autonomous social publishing infrastructure that gives OpenClaw agents a single secure API to draft, schedule, and publish content across multiple channels.
Featured*
