fact-checker-mcp

Joopsnijder/fact-checker-mcp

3.2

If you are the rightful owner of fact-checker-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The Fact Checker MCP Server is a hybrid implementation combining CrewAI multi-agent system and MCP server technology for robust fact-checking.

Tools
4
Resources
0
Prompts
0

Fact Checker MCP Server

Hybrid CrewAI + MCP implementation for fact-checking with multi-agent verification system.

Overview

This tool works as both:

  1. CrewAI multi-agent system for comprehensive fact-checking with multiple specialized agents
  2. MCP server for direct integration with Claude Desktop and other MCP clients

Installation & Development

This project uses UV for dependency management and virtual environment handling.

Setup

  1. Install dependencies:

    uv sync

  2. Environment variables: Create a .env file with:

    OPENAI_API_KEY=your_openai_api_key
SERPER_API_KEY=your_serper_api_key  # Optional: for enhanced web search
BRAVE_API_KEY=your_brave_api_key    # Optional: for additional search provider

Usage

As MCP Server (for Claude Desktop)

Development mode:

mcp dev fact-checker.py

Production mode:

uv run python fact-checker.py --mcp

As Standalone Application

Interactive mode:

uv run python fact-checker.py --check

File input:

uv run python fact-checker.py --check input_text.txt

Pipeline usage:

echo "Your text to fact-check" | uv run python fact-checker.py --check

In Python Code

from fact_checker import run_fact_check_crew

# Analyze text for fact checking
report = run_fact_check_crew("Your text to analyze")
print(report.overall_reliability)

Help

Display usage instructions:

python fact-checker.py --help

MCP Server Tools

When running as MCP server, the following tools are available:

  • quick_verify(text: str) - Fast verification of short claims (< 500 chars)
  • deep_fact_check(text: str) - Comprehensive multi-agent fact-checking
  • check_specific_statistic(statistic: str, context: str, year: int) - Targeted statistic verification
  • get_history_summary() - Summary of all completed fact checks

MCP Resources

  • history://list - List of all fact-check reports
  • history://report/{id} - Specific fact-check report details

Features

  • Multi-Provider Search: Automatic fallback between Serper, SearXNG, Brave Search, and web scraping
  • Smart Usage Tracking: Monitors API usage limits and rotates between providers
  • Comprehensive Analysis: Extracts and verifies claims, statistics, quotes, and facts
  • Detailed Reporting: Provides confidence scores, sources, and explanations
  • Memory & Caching: Built-in caching and memory management for efficiency

Code Quality

Format and lint:

uv run ruff format .
uv run ruff check . --fix

Run tests:

uv run pytest

Architecture

  • fact-checker.py - Main server with CrewAI agents and MCP endpoints
  • smart_search_tool.py - Multi-provider search system with automatic fallback
  • pyproject.toml - UV project configuration with all dependencies

Requirements

  • Python 3.10+
  • OpenAI API key (required)
  • Serper API key (optional, enhances search capabilities)
  • Brave API key (optional, additional search provider)

Virtual Environment

The project uses UV's managed virtual environment. The previous manual .venv setup has been replaced with proper UV project structure to resolve MCP CLI compatibility issues.