# Extracta MCP for AI Agents AI Agent Connect

> Extracta automates the process of turning messy documents like PDFs and images into clean, structured JSON data. It handles complex AI classification for things like receipts and invoices, letting you build custom extraction schemas on the fly. You can manage your entire document pipeline, from submitting URLs to auditing historical results, all through your preferred AI agent.

## Overview
- **Category:** artificial-intelligence
- **Price:** Free
- **Endpoint:** https://edge.vinkius.com/vk_preview_C65ttiBqS65NCFErwJyenYxrCFQRkfcwF5fUOGUO/ai-agent-connect
- **Tags:** ocr, data-extraction, document-classification, json-parsing, automated-data-entry, unstructured-data

## Description

Extracta changes how you handle document data. Instead of your team manually opening PDFs, highlighting text, and copy-pasting values into a spreadsheet, you can now point your AI agent at a list of URLs. The Connector does the heavy lifting of reading the files, identifying the important information based on the rules you set, and handing you back clean JSON data ready for your database. It handles the nuance of document types too, so you can tell it to sort files into categories like invoices or contracts before you even look at them. This setup is a major shift from looking at files to receiving data. Because it's available in the Vinkius catalog, you can plug it into your existing agent workflow without worrying about the infrastructure. You can tweak your extraction rules as your business needs change, adding new fields or adjusting how data is mapped without starting over. It turns a tedious manual chore into a background process that just works.

## Tools

### get_batch_results
Pull a bulk list of historical results from a previous extraction run. This helps you audit your data in one go.

### create_classification
Create a new setup to sort documents into types like invoices or contracts. Pass a JSON schema to define your categories.

### get_classification_results
Find out what category the AI assigned to a specific document. Use this to verify your sorting rules.

### create_extraction
Set up a new data extraction process with custom JSON fields and rules. It returns an ID for future use.

### delete_extraction
Remove a specific extraction process so it can no longer receive files. This is useful for cleaning up old configurations.

### get_results
Check the current status or the final data for a single document. Use this to see if a file is done processing.

### update_extraction
Modify the mapping rules of an existing process to refine how data is parsed. You don't have to create a new endpoint to do this.

### upload_file_url
Submit a public URL for a document to begin the asynchronous extraction process. This returns a document ID for tracking.

### view_extraction
Review the configuration and fields of an existing extraction process. Check your settings at any time.

### view_classification
View the specific details and rules of an existing document classification process.

## Prompt Examples

**Prompt:** 
```
I have a bunch of receipts in this folder. Can you extract the dates and totals for all of them?
```

**Response:** 
```
I've processed the receipts for you. Here is the structured data:

| Date | Vendor | Total |
| :--- | :--- | :--- |
| 2023-10-01 | Starbucks | $12.50 |
| 2023-10-02 | Shell | $45.00 |
| 2023-10-05 | Whole Foods | $88.20 |

All totals have been extracted into the JSON format you requested.
```

**Prompt:** 
```
Create a process to pull the vendor name and tax amount from my monthly invoices.
```

**Response:** 
```
I've set up a new extraction process for your invoices.

**Configuration Details:**
- **Fields:** vendor_name, tax_amount
- **Format:** JSON
- **Process ID:** ext_9921

You can now start providing URLs for your invoices, and I'll extract those specific details for you.
```

**Prompt:** 
```
What category does this document fall into?
```

**Response:** 
```
I've analyzed the document you provided.

**Classification Result:**
- **Type:** Invoice
- **Confidence Score:** 98%
- **Reasoning:** The document contains standard billing headers, a line item table, and a clear total amount, which matches your invoice criteria.
```

## Capabilities

### Create custom extraction schemas
Define specific fields like dates, amounts, and item descriptions for your documents.

### Process files via URLs
Submit links to PDFs or images to trigger automatic data gathering.

### Sort documents into categories
Automatically label files as invoices, receipts, or contracts based on AI predictions.

### Check extraction status
Monitor how the AI is doing with a specific file in real-time.

### Pull historical batch results
Get a list of every document you've processed in the past to track your history.

### Edit existing extraction rules
Update your data mapping without having to create new endpoints.

## Use Cases

### Automated Invoice Processing
An operations lead asks the agent to extract totals from 50 different invoice URLs and save them to a list.

### Document Sorting
A finance person wants the agent to look at a folder of uploads and tell them which ones are contracts.

### Data Migration
A developer uses the tool to turn a pile of old JPG receipts into a structured JSON database.

### Audit Trail
A compliance officer asks the agent to pull the last 100 extraction results to verify accuracy.

## Benefits

- Stop manual data entry by using the creation tools to map fields like totals and dates automatically.
- Organize your files instantly with the classification setup to separate contracts from receipts.
- Audit your history easily with the batch results tool to see everything you've processed in one place.
- Update your logic on the fly with the update tool without needing to rebuild your entire pipeline.
- Track real-time progress with the results tool to see exactly when your data is ready for use.
- Save time on file handling by using the URL upload tool to process documents directly from the web.

## How It Works

The bottom line is you stop manually typing data from documents and start receiving clean JSON automatically.

1. Connect your Extracta.ai API key to the Connector.
2. Define your JSON schemas and classification rules via natural conversation.
3. Provide document URLs to your agent and receive structured JSON.

## Frequently Asked Questions

**Can Extracta handle different file types?**
Extracta supports common file formats including PDFs, JPGs, and PNGs. This means you can send images of physical receipts or digital PDFs for processing.

**How does Extracta help with invoice processing?**
It automates the extraction of specific fields like dates, vendor names, and totals from invoices. It turns those documents into structured JSON data for your records.

**Can I change the fields I'm extracting later?**
Yes, you can update your extraction rules at any time. You can add or remove fields like tax amounts or item descriptions without having to rebuild your setup.

**Does Extracta work with my current AI client?**
Yes, it works with Claude, Cursor, Windsurf, and other MCP-compatible clients. You can manage your document workflows directly within the tools you already use.

**How do I see my previous results?**
You can fetch a list of all previously processed documents and their data payloads. This is great for auditing your history or checking past extractions.

**Can it tell the difference between a receipt and a contract?**
Yes, it uses AI classification to automatically sort documents into categories. You can define these types, and the agent will sort them for you based on the content.

**Do I need to manually upload every file?**
No, you can submit publicly accessible URLs for documents. The agent will then trigger the extraction workflow automatically.

**Can my agent create a new data extraction setup with custom fields?**
Yes. Use the 'create_extraction' tool. Provide a JSON schema defining the fields you expect (e.g., 'total_amount', 'vendor_name'). The agent will return a new extractionId for document processing.

**How do I process a PDF document using a specific extraction ID via chat?**
Use the 'upload_file_url' tool. Provide the extractionId and the public URL of your PDF. The agent will trigger the workflow and return a documentId, which you can use with 'get_results' to fetch the data.

**Can I see the predicted document type and confidence score through the agent?**
Absolutely. Use the 'get_classification_results' tool with the document and classification IDs. The agent will retrieve the AI-predicted label (e.g., 'Invoice') and the confidence score for the processed file.