# JigsawStack MCP

> JigsawStack connects your AI client to a utility belt of specialized APIs for data extraction and content analysis. Use it to scrape structured data from any website, translate text between languages, summarize long documents, or convert addresses into coordinates—all through natural conversation.

## Overview
- **Category:** developer-tools
- **Price:** Free
- **Tags:** web-scraping, text-to-sql, ocr, data-validation, geocoding, api-utility

## Description

This MCP lets you treat complex data tasks like simple chat commands. Instead of jumping between ten different services, your agent handles everything in one place. You can pull structured product lists from e-commerce sites using `ai_web_scrape`, then send the text to another utility to validate all the associated email addresses. Need to know where an old address is? Just ask it to geocode the location and convert that into coordinates. If you've got a massive chunk of technical text, use the summarization tool first, then pass the resulting short copy through the translation utility if you need it in Spanish or French. All these capabilities are managed under one roof at Vinkius, making your AI agent a full-stack data worker.

## Tools

### ai_web_scrape
Automatically pulls structured data from a given website address using natural language prompts.

### check_nsfw_image
Determines if an uploaded image contains inappropriate or NSFW content.

### check_profanity_text
Scans a provided block of text and flags any instances of profane language.

### summarize_text
Condenses long articles, reports, or documents into key takeaways while preserving core meaning.

### translate_text
Converts text content accurately between multiple foreign languages.

### extract_text_from_image
Performs OCR on an image URL, pulling out readable text content from photos or scans.

### ai_web_search
Runs a comprehensive, AI-powered search across the web to gather current information and sources.

## Prompt Examples

**Prompt:** 
```
Scrape the top 10 products from this e-commerce page and validate the contact emails.
```

**Response:** 
```
Scraped 10 products from example-shop.com. 1) 'Wireless Headphones' ($79.99, 4.5★). 2) 'Smart Watch' ($199.99, 4.8★). 3) 'USB-C Hub' ($34.99, 4.2★). Emails validated: contact@shop.com ✅ valid, support@shop.com ✅ valid, info@shop.com ⚠️ catch-all.
```

**Prompt:** 
```
Generate a SQL query to find the top 5 customers by revenue and geocode their addresses.
```

**Response:** 
```
SQL generated: `SELECT c.name, SUM(o.total) as revenue FROM customers c JOIN orders o ON c.id = o.customer_id GROUP BY c.id ORDER BY revenue DESC LIMIT 5`. Geocoded: 1) '123 Main St, NYC' → 40.7128, -74.0060. 2) '456 Market St, SF' → 37.7749, -122.4194. 3) '789 Michigan Ave, Chicago' → 41.8781, -87.6298.
```

**Prompt:** 
```
Translate this product description to Spanish, French, and Portuguese.
```

**Response:** 
```
Translations complete! ES: 'Transforma tus datos en insights accionables con nuestra plataforma de analytics.' FR: 'Transformez vos données en insights actionnables avec notre plateforme d'analytics.' PT: 'Transforme seus dados em insights acionáveis com nossa plataforma de analytics.' All 3 translations delivered with natural phrasing.
```

## Capabilities

### Extracting web content
You instruct your agent to scrape specific types of information from any public website.

### Analyzing text and images
You submit files or text for analysis, letting the system check for profanity, summarize long passages, or identify NSFW content in an image.

### Translating languages
You pass text to translate between multiple supported languages instantly.

### Geolocating addresses
You provide a physical address, and the system converts it into precise latitude and longitude coordinates.

### Generating web search results
You ask your agent to perform an AI-powered search across the current web for up-to-date information.

## Use Cases

### Analyzing Competitor Pricing
A researcher needs to know what a competitor is selling and how much it costs globally. They ask their agent to use `ai_web_scrape` on the site, then pass the results through `translate_text` to get pricing data in Euros for an international report.

### Cleaning Up Form Submissions
An operations team receives web forms with user-submitted addresses. They use the MCP's geocoding capabilities and validation utilities simultaneously, ensuring every address is clean and mapped to precise coordinates before entering it into the CRM.

### Preparing a Global Report
A consultant has a 50-page whitepaper. They use `summarize_text` to cut it down to an executive summary, then use `translate_text` on that short version to deliver the core message to stakeholders in Mandarin.

### Monitoring Content Safety
A moderator needs to check a batch of user-submitted forum posts. They send the text through `check_profanity_text` and any attached images through `check_nsfw_image`, ensuring immediate content compliance before review.

## Benefits

- You bypass manual data entry by letting `ai_web_scrape` extract structured product tables and lists from any public website. This saves hours of copy-pasting into spreadsheets.
- Need to check content compliance? Send text through the profanity checker or use `check_nsfw_image` on uploaded media, ensuring all outgoing material meets brand guidelines before publication.
- Never worry about language barriers again. The `translate_text` tool gives you accurate translations instantly, allowing your agent to localize marketing copy for dozens of markets in minutes.
- When you hit a dead end with old documents or photos, use the OCR utility (`extract_text_from_image`) to turn images into usable text that can then be summarized or validated.
- Instead of performing multiple API calls yourself, this MCP chains them together. Your agent handles the flow: scrape -> validate -> translate in one prompt.

## How It Works

The bottom line is you talk naturally, and your agent executes a complex chain of utilities behind the scenes.

1. Subscribe to this MCP and enter your JigsawStack API Key into your preferred AI client.
2. Tell your agent what you need. For example, 'Scrape the top 10 items from this page and translate the titles.'
3. The system runs the necessary tools—like `ai_web_scrape` followed by `translate_text`—and presents the finished data back to you.

## Frequently Asked Questions

**Can JigsawStack MCP handle scraping multiple pages?**
Yes, you can chain scraping actions together. You might use `ai_web_scrape` on a main index page and then ask the agent to iterate through links found there for more detailed product data.

**Does JigsawStack MCP just scrape text?**
No, it's much broader. It also includes tools like `check_nsfw_image` and `extract_text_from_image`, letting you process images directly for content.

**How does the JigsawStack MCP handle data validation?**
The MCP has built-in utilities that validate common formats, like emails and phone numbers. This ensures that when your agent scrapes a list of contacts, you only get usable records.

**Can I use JigsawStack MCP for data pipelines?**
Absolutely. You can create complex multi-step workflows—for example, scrape text, summarize it, and then translate the summary into a target language all in one command sequence.

**Is translation limited to specific languages with JigsawStack MCP?**
The `translate_text` tool supports multiple languages. Just specify the source and target languages you need when giving the instruction to your agent.