agent-farm 0.1.3

DuckDB-powered MCP Server with SQL macros for LLM agents

Agent Farm

DuckDB-powered MCP Server with SQL macros for LLM agents - Web Search, Python execution, RAG, and more.

DuckDB - Ollama - Docker - Query Farm

Features

• MCP Server: Exposes DuckDB as an MCP server for Claude and other LLM clients
• Auto-Discovery: Automatically discovers MCP configurations from standard locations
• LLM Integration: SQL macros for calling Ollama models (local and cloud)
• Tool Calling: Full function calling support for agentic workflows
• Web Search: DuckDuckGo and Brave Search integration
• Shell Execution: Run shell commands and Python code via UV
• Web Scraping: Fetch and extract text from web pages
• RAG Support: Embeddings and vector similarity search
• Rich Extensions: Pre-configured with useful DuckDB community extensions

Installation

Using uv (recommended): uv sync --dev

Or with pip: pip install -e .

Quick Start

Run the MCP server: agent-farm

Or as a module: python -m agent_farm

SQL Macros

Cloud LLM Models (via Ollama)

SELECT deepseek('Explain quantum computing'); SELECT kimi_think('Solve this step by step: ...'); SELECT qwen3_coder('Write a Python function for...');

Web Search

SELECT ddg_instant('Python programming'); SELECT ddg_abstract('machine learning'); SELECT brave_search('DuckDB tutorial');

Shell and Python Execution

SELECT shell('ls -la'); SELECT py('print(2+2)'); SELECT py_with('requests', 'import requests; print(requests.version)');

Web Scraping

SELECT fetch('https://example.com'); SELECT fetch_text('https://example.com'); SELECT fetch_json('https://api.example.com/data');

File and Git Operations

SELECT read_file('path/to/file.txt'); SELECT git_status(); SELECT git_log(10);

RAG and Embeddings

SELECT rag_query('What is the price?', 'Product: Widget, Price: 49.99'); SELECT embed('Hello world'); SELECT semantic_score('query', 'document');

Combined Power Macros

SELECT search_and_summarize('What is DuckDB?'); SELECT analyze_page('https://example.com', 'What is this page about?'); SELECT review_code('src/main.py');

Docker

docker build -t agent-farm . docker run -it agent-farm

Requirements

• Python >= 3.11
• DuckDB >= 1.1.0
• Ollama (for LLM features)

License

MIT