Case Study Generator MCP Server

A Model Context Protocol (MCP) server that processes document content and GitHub repositories with Gemma3 to extract structured business insights for case study generation.

Overview

This MCP server provides three main capabilities:

1. Document Processing - Extract business insights from documents (proposals, case studies, contracts)
2. GitHub Repository Analysis - Analyze repositories for business value and technical benefits
3. Company Research - Real-time web research using Tavily + AI analysis for company insights

The server uses Gemma3 8B-Instruct via Ollama for local LLM processing, ensuring privacy and control over your data.

Architecture

User → Claude Desktop → Google Drive (retrieve) → MCP Server (process) → Claude Desktop (reason & write) → Google Drive (save)

• Claude Desktop: Handles document retrieval, reasoning, writing, and saving
• MCP Server: Processes data with Gemma3 and returns structured insights
• Ollama/Gemma3: Local LLM for business analysis and insight extraction

Prerequisites

Required Software

1. Python 3.11+ - Programming language runtime
2. Ollama - Local LLM inference server
3. Gemma3 Model - Language model for analysis

Install Ollama

Visit ollama.ai and install for your platform.

After installation, pull the Gemma3 model:

ollama pull gemma3n:e4b

Verify Ollama is running:

ollama list

Installation

Option 1: Using venv (Recommended)

# Clone the repository
git clone <repository-url>
cd case-study-mcp

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Option 2: Using Poetry

# Clone the repository
git clone <repository-url>
cd case-study-mcp

# Install with Poetry
poetry install
poetry shell

Configuration

Environment Variables (Optional)

Create a .env file in the project root:

# GitHub API token for higher rate limits (optional)
GITHUB_TOKEN=your_github_token_here

# Tavily API key for web search company research (optional)
TAVILY_API_KEY=your_tavily_api_key_here

# Ollama configuration
OLLAMA_HOST=http://localhost:11434
OLLAMA_MODEL=gemma3n:e4b

# Logging level
LOG_LEVEL=INFO

GitHub Token Setup

For better GitHub API rate limits, create a personal access token:

1. Go to GitHub Settings → Developer settings → Personal access tokens
2. Generate a new token with public_repo scope
3. Add to .env file or set as environment variable

Tavily API Setup (For Company Research)

For real company research capabilities, get a Tavily API key:

1. Sign up at tavily.com
2. Get your API key from the dashboard
3. Add TAVILY_API_KEY=your_key_here to .env file

Note: Without Tavily, company research will use LLM pattern matching only.

Usage

Starting the MCP Server

# Run the server
python mcp_server.py

The server communicates via stdio and will wait for MCP protocol messages.

Integration with Claude Desktop

Add to your Claude Desktop MCP configuration:

{
  "mcpServers": {
    "case-study-mcp": {
      "command": "python",
      "args": ["/path/to/case-study-mcp/mcp_server.py"],
      "cwd": "/path/to/case-study-mcp"
    }
  }
}

Example Usage in Claude Desktop

User: "Create a case study for TechCorp using github.com/example/ecommerce-platform and my retail proposal document"

Claude: "I'll analyze the repository and document to create your case study..."

[Claude uses MCP tools:]
1. process_document_text(text="retail proposal content", doc_type="proposal")
2. analyze_github_repo(repo_url="github.com/example/ecommerce-platform") 
3. research_company_basic(company_name="TechCorp")

[Claude synthesizes results into compelling case study]

MCP Tools

1. process_document_text

Extract business insights from document content.

Parameters:

• text (required): Document content text
• doc_type (optional): Type of document - "proposal", "case_study", "contract", or "general"

Returns:

{
  "success": true,
  "document_type": "proposal",
  "insights": {
    "challenges": ["Legacy platform scalability issues"],
    "solutions": ["Microservices architecture"],
    "metrics": ["90% improvement in load times"],
    "context": "Mid-market e-commerce company",
    "key_stakeholders": ["CTO", "Engineering Team"],
    "timeline": "6-month implementation"
  }
}

2. analyze_github_repo

Analyze GitHub repository for business value.

Parameters:

• repo_url (required): GitHub repository URL

Returns:

{
  "success": true,
  "repository": {
    "url": "github.com/example/repo",
    "name": "example/ecommerce-platform",
    "tech_stack": ["Python", "React", "Docker"],
    "stars": 1250,
    "language": "Python"
  },
  "business_analysis": {
    "problem_solved": "Scalable e-commerce platform",
    "key_features": ["Microservices", "Redis caching"],
    "business_value": "Enables 10x traffic scaling",
    "technical_benefits": ["Horizontal scaling", "Fault tolerance"],
    "target_users": "Mid to large-scale e-commerce businesses",
    "scalability": "Designed for high-traffic peaks",
    "integration_points": ["CRM systems", "Payment gateways"]
  }
}

3. research_company_basic

Real company research using web search + AI analysis.

Parameters:

• company_name (required): Name of the company
• company_context (optional): Additional context about the company

Returns:

{
  "success": true,
  "company": {
    "name": "TechCorp",
    "profile": "Technology company specializing in...",
    "industry": "Technology",
    "business_model": "SaaS"
  },
  "insights": {
    "challenges": ["Digital transformation needs"],
    "opportunities": ["Cloud migration potential"],
    "technology_needs": ["Scalable infrastructure"]
  },
  "web_search_used": true,
  "search_sources": [
    {"title": "TechCorp - About Us", "url": "https://techcorp.com/about"},
    {"title": "TechCorp on LinkedIn", "url": "https://linkedin.com/company/techcorp"}
  ]
}

Testing

Manual Testing

Test each tool individually:

# Test document processing
echo '{"text": "Sample proposal text", "doc_type": "proposal"}' | python test_document.py

# Test GitHub analysis  
echo '{"repo_url": "github.com/microsoft/vscode"}' | python test_github.py

# Test company research
echo '{"company_name": "Microsoft"}' | python test_company.py

Health Check

The server provides a health check resource:

URI: health://status

Returns status of all components including Gemma3, GitHub API, and processors.

Troubleshooting

Common Issues

1. Ollama Connection Error

Error: Failed to connect to Ollama

Solution: Ensure Ollama is running (ollama serve) and the model is pulled (ollama pull gemma3n:e4b).

2. GitHub Rate Limit

Error: GitHub API rate limit exceeded

Solution: Add a GitHub token to your .env file for higher limits.

3. Model Not Found

Error: Model ollama3n:4b not found

Solution: Pull the model with ollama pull ollama3n:4b.

4. Import Errors

ImportError: No module named 'mcp'

Solution: Install dependencies with pip install -r requirements.txt.

5. Company Research Limited

Warning: Tavily not configured. Company research will use LLM patterns only.

Solution: Get a Tavily API key from tavily.com and add to .env file.

Performance Optimization

• Memory Usage: Ollama3n 4B requires ~4-6GB RAM for optimal performance
• Processing Time: Document processing typically takes 5-15 seconds
• Concurrent Requests: Server handles one request at a time by design

Logging

Enable debug logging:

export LOG_LEVEL=DEBUG
python mcp_server.py

Project Structure

case-study-mcp/
├── mcp_server.py              # Main MCP server entry point
├── gemma3_client.py           # Gemma3/Ollama client wrapper
├── document_processor.py      # Document analysis processor
├── github_analyzer.py         # GitHub repository analyzer  
├── company_researcher.py      # Company research processor
├── prompts.py                 # Gemma3 prompt templates
├── pyproject.toml            # Poetry configuration
├── requirements.txt          # Pip dependencies
├── README.md                 # This file
├── project_config.md         # Project specification
└── workflow_state.md         # Development workflow tracking

Development

Contributing

1. Fork the repository
2. Create a feature branch
3. Make changes with tests
4. Submit a pull request

Code Style

• Use Black for formatting: black .
• Use isort for imports: isort .
• Use mypy for type checking: mypy .

License

MIT License - see LICENSE file for details.

Support

For issues and questions:

1. Check the troubleshooting section above
2. Review the project configuration in project_config.md
3. Open an issue with detailed error logs and steps to reproduce