AnyDB MCP Server

AnyDB MCP Server

A Model Context Protocol (MCP) server that provides intelligent database operations through natural language processing. This server integrates SQLite databases with Ollama for AI-powered SQL generation and execution.

Features

Core Database Operations

• Natural Language to SQL: Convert plain English instructions into SQL queries using Ollama
• Universal Database Operations: Works with any SQLite table/entity without predefined schemas
• Automatic Schema Evolution: Dynamically adds columns to tables when new data requires them
• MCP Integration: Seamlessly integrates with Claude Desktop and other MCP-compatible clients
• Async Operations: Built on modern Python async/await for high performance
• Safety First: Separate tools for read and write operations

Vector Database & RAG

• File Embedding: Automatically convert files into vector embeddings for semantic search
• Semantic Search: Find relevant content using natural language queries instead of exact keyword matching
• RAG Support: Enable Claude Desktop to answer questions about uploaded documents with context
• Smart Chunking: Intelligently splits large documents into overlapping chunks for better retrieval
• Persistent Storage: ChromaDB-powered vector database with automatic embedding generation

Web Scraping & Knowledge Base (NEW!)

• URL Scraping: Extract and store content from web pages automatically
• Smart Content Extraction: Clean HTML and extract meaningful text with metadata
• URL Security: Built-in validation and security checks to prevent malicious URLs
• Semantic Web Search: Query scraped web content using natural language
• Web Knowledge Management: List, search, and manage scraped web pages

Available Tools

Database Tools

1. query_entity

Query any table with natural language instructions.

Parameters:

• entity_name (required): Name of the table to query
• instruction (optional): Natural language query instruction (default: "SELECT all records")

Example: Query users table for active accounts

2. insert_entity

Insert records into any table using natural language descriptions. Automatically adds new columns if the data contains fields not present in the current table schema.

Parameters:

• entity_name (required): Name of the table
• data (required): Data to insert (JSON or natural description)

Example: Insert a new user with email and name Auto-Schema: If inserting "user with name John, email john@test.com, and premium status" into a table that only has name/email columns, the system will automatically add a "status" column

3. update_entity

Update records in any table with conditions. Automatically adds new columns if the update introduces new fields.

Parameters:

• entity_name (required): Name of the table
• instruction (required): Update instruction
• conditions (optional): WHERE conditions

Example: Update user status to active where email matches Auto-Schema: If updating with "set premium status to gold and loyalty points to 100" on a table without these columns, they will be added automatically

4. delete_entity

Delete records from any table with optional conditions.

Parameters:

• entity_name (required): Name of the table
• conditions (optional): WHERE conditions for deletion

Example: Delete inactive users older than 30 days

5. create_table

Create new tables with AI-generated schemas.

Parameters:

• entity_name (required): Name of the new table
• schema_description (required): Description of table schema

Example: Create a products table with name, price, and category

6. sql_query

Execute raw SQL SELECT queries directly.

Parameters:

• query (required): SQL query to execute

Example: Direct SQL for complex joins and analytics

7. sql_execute

Execute raw SQL modification queries (INSERT, UPDATE, DELETE, CREATE, etc.).

Parameters:

• query (required): SQL query to execute

Example: Direct SQL for complex data modifications

Vector Database Tools (NEW!)

8. add_file_to_vector_db

Add a file to the vector database for semantic search and RAG (Retrieval Augmented Generation).

Parameters:

• filename (required): Name of the file
• content (required): Content of the file (text)
• metadata (optional): Optional metadata for the file

Example: Add a document about machine learning for later semantic search

9. search_vector_db

Search the vector database for relevant file content using semantic similarity.

Parameters:

• query (required): Search query for semantic similarity
• max_results (optional): Maximum number of results to return (default: 5)

Example: Find documents related to "neural networks and AI"

10. list_vector_files

List all files stored in the vector database.

Parameters: None

Example: View all documents available for search

11. remove_file_from_vector_db

Remove a file from the vector database.

Parameters:

• filename (required): Name of the file to remove

Example: Delete outdated documents from the knowledge base

Web Scraping Tools (NEW!)

12. scrape_url

Scrape content from a web page and store it in the vector database for semantic search and RAG.

Parameters:

• url (required): URL of the web page to scrape
• custom_filename (optional): Custom filename for the scraped content

Example: Scrape a Wikipedia article or documentation page for later querying

13. query_web_content

Query scraped web page content using semantic search to find relevant information.

Parameters:

• query (required): Search query for finding relevant web content
• max_results (optional): Maximum number of results to return (default: 5)

Example: Search scraped web pages for information about specific topics

14. list_scraped_pages

List all scraped web pages stored in the vector database with metadata.

Parameters: None

Example: View all websites you've scraped and stored

15. remove_scraped_page

Remove a scraped web page from the vector database.

Parameters:

• filename (required): Filename of the scraped page to remove

Example: Clean up outdated or unwanted scraped web content

Installation

Prerequisites

• Python 3.8+
• Ollama running locally
• Claude Desktop (for MCP integration)

Setup

1. Clone the repository:

2. Install dependencies:

3. Start Ollama:

4. Run the server:

Claude Desktop Integration

Add this server to Claude Desktop by editing your config file:

Windows: %APPDATA%\Claude\claude_desktop_config.json

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Restart Claude Desktop to connect the server.

Configuration

Ollama Settings

Default configuration in mcp_server.py:

• Host: localhost
• Port: 1434
• Model: llama3.1

Database Settings

• Default DB: anydb.sqlite (created automatically)
• Location: Same directory as the server
• Type: SQLite with foreign key constraints enabled

Usage Examples

Once integrated with Claude Desktop, you can use natural language:

Database Operations

• "Create a users table with id, name, email, and created_at fields"
• "Show me all active users from the last 30 days"
• "Insert a new product: iPhone 15, price $999, category Electronics"
• "Update all pending orders to processed where amount > 100"
• "Delete test users where email contains 'test'"

Smart Schema Evolution

• "Insert a book: Title 'AI Handbook', Author 'Jane Doe', ISBN '123456', Format 'Hardcover'" (automatically adds Format column if missing)
• "Add employee with name John, salary 75000, department IT, and remote status" (adds department and remote columns as needed)
• "Update product with warranty period 2 years and eco-friendly rating A+" (dynamically expands product schema)

Vector Database & File Operations

• "Add this document to the knowledge base" (when attaching a file in Claude Desktop)
• "Search for information about machine learning algorithms"
• "Find documents related to user authentication and security"
• "What does the uploaded contract say about payment terms?"
• "Show me all documents I've added to the database"
• "Remove the old privacy policy document"

Web Scraping & Knowledge Base

• "Scrape this Wikipedia article about neural networks: https://en.wikipedia.org/wiki/Neural_network"
• "Save the content from this blog post for later reference"
• "What information did I scrape about machine learning from that research paper website?"
• "Search my scraped web pages for information about Python best practices"
• "Show me all the web pages I've scraped and stored"
• "Remove the outdated documentation I scraped last month"

Architecture

Development

Project Structure

Key Components

Core Modules:

• main.py: Entry point with dependency checking and startup information
• mcp_server.py: MCP protocol implementation, tool registration, and request routing
• dbtool.py: Database operations, SQL generation, and data management
• filetool.py: Vector database operations, file processing, and semantic search
• webscrapertool.py: Web scraping, content extraction, and URL processing

Business Logic Classes:

• DatabaseManager: Handles async SQLite operations and database connections
• DatabaseTools: High-level database operations with natural language support
• OllamaClient: Manages AI model communication for SQL generation
• VectorDatabaseManager: Manages ChromaDB operations and document embeddings
• FileTools: High-level file operations and semantic search functionality
• WebScraperManager: Handles web page fetching, content extraction, and URL validation
• WebScraperTools: High-level web scraping operations with vector database integration

Troubleshooting

Common Issues

1. Server won't start: Check if Ollama is running on port 1434
2. No tools showing in Claude: Verify MCP config path and restart Claude Desktop
3. SQL errors: Check table names and ensure proper natural language descriptions
4. Ollama connection failed: Confirm Ollama model is installed and accessible

Debug Mode

Run with Python's verbose mode for detailed logs:

License

This project is open source. See LICENSE file for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

Support

For issues and questions:

• Check the troubleshooting section
• Review Ollama and MCP documentation
• Open an issue on the repository