🚀 Gemini + CrewAI + MCP Server Setup Guide

This guide walks you through setting up the JIRA report agent using Google Gemini as the LLM with CrewAI framework and your existing jira-mcp-snowflake MCP server.

🎯 Architecture Overview

Gemini LLM ← CrewAI Agents ← MCP Tools → jira-mcp-snowflake Server → JIRA Data

📋 Prerequisites

1. Google Gemini API Key - Get from Google AI Studio
2. Your existing MCP server - The jira-mcp-snowflake server we tested
3. Python 3.9+ with pip

⚡ Quick Start

1. Install Dependencies

# Install the Gemini + CrewAI requirements
pip install -r requirements.txt

2. Environment Setup

Create a .env file or set environment variables:

# Copy the example and edit with your values
cp gemini_env_example.txt .env

# Edit .env with your actual API key:
# GEMINI_API_KEY=your-actual-gemini-api-key-here
# MCP_SERVER_URL=http://localhost:8000

Or export directly:

export GEMINI_API_KEY="your-gemini-api-key-here"
export MCP_SERVER_URL="http://localhost:8000"  # Your MCP server URL

3. Run the Gemini Agent

python3 crewai_gemini.py

🔧 Configuration Options

Gemini Model Selection

Choose between different Gemini models based on your needs:

# Fast and cost-effective (default)
export GEMINI_MODEL="gemini-1.5-flash"

# More capable for complex analysis
export GEMINI_MODEL="gemini-1.5-pro"

LLM Provider Switch

You can switch between Gemini and OpenAI:

# Use Gemini (default)
export LLM_PROVIDER="gemini"
export GEMINI_API_KEY="your-key"

# Use OpenAI instead
export LLM_PROVIDER="openai"
export OPENAI_API_KEY="your-key"

MCP Server Configuration

# Local MCP server (default)
export MCP_SERVER_URL="http://localhost:8000"

# Remote MCP server
export MCP_SERVER_URL="https://your-mcp-server.com"

# Disable MCP and use direct JIRA API
export USE_MCP_SERVER="false"
export JIRA_BASE_URL="https://your-jira.com"
export JIRA_USERNAME="your-username"
export JIRA_API_TOKEN="your-token"

🎭 Agent Architecture

The Gemini implementation uses 3 specialized agents:

1. 🔍 JIRA Data Specialist

• Role: Collects data from MCP server
• Tools: MCPJIRADataTool
• Focus: Efficient data extraction from projects

2. 📊 Data Analysis Expert

• Role: Filters and validates data
• Tools: DateFilterTool
• Focus: Temporal analysis and data quality

3. 📝 Technical Report Specialist

• Role: Creates professional reports
• Tools: None (pure LLM reasoning)
• Focus: Clear stakeholder communication

🔄 Workflow Process

graph TD
    A[Start] --> B[Data Collector Agent]
    B --> C[Fetch JIRA issues via MCP]
    C --> D[Data Analyst Agent]
    D --> E[Filter last 7 days]
    E --> F[Report Generator Agent]
    F --> G[Create markdown report]
    G --> H[weekly_jira_report_gemini.md]

Loading

✅ Verification Steps

1. Test Environment

python crewai_config.py

Expected output:

✅ Environment configured for GEMINI + MCP integration
- LLM Provider: gemini
- Using MCP: true
- Projects: ['CCITJEN', 'CCITRP', 'QEHS']
- MCP URL: http://localhost:8000
- LLM created successfully: ChatGoogleGenerativeAI

2. Test MCP Connection

Verify your MCP server is accessible:

curl -X POST http://localhost:8000/mcp/jira-mcp-snowflake/list_jira_issues \
  -H "Content-Type: application/json" \
  -d '{"project": "CCITJEN", "status": "6", "limit": 5}'

3. Run Full Agent

python crewai_gemini_implementation.py

🚨 Troubleshooting

Common Issues

1. Gemini API Key Error

ValueError: GEMINI_API_KEY environment variable is required

Solution:

• Get API key from Google AI Studio
• Set environment variable: export GEMINI_API_KEY="your-key"

2. MCP Server Connection Failed

Failed to connect to MCP server: Connection refused

Solutions:

• Check if MCP server is running: curl http://localhost:8000/health
• Verify URL in MCP_SERVER_URL environment variable
• Check firewall/network connectivity

3. Rate Limiting

429 Too Many Requests

Solutions:

• Switch to gemini-1.5-flash for higher rate limits
• Add delays between requests
• Upgrade Gemini API quota

4. Tool Import Errors

ModuleNotFoundError: No module named 'langchain_google_genai'

Solution:

pip install -r requirements_crewai.txt

📊 Performance Comparison

Aspect	Gemini 1.5 Flash	Gemini 1.5 Pro
💰 Cost	Lower	Higher
⚡ Speed	Faster	Slower
🧠 Capability	Good for reports	Better for analysis
📊 Complex Reasoning	Basic	Advanced
📝 Report Quality	High	Very High

🎯 Best Practices

1. Model Selection

• Use gemini-1.5-flash for regular reporting (default)
• Use gemini-1.5-pro for complex analysis or when quality is critical

2. Error Handling

• The implementation includes robust error handling for MCP failures
• Automatic fallback to error messages in reports
• Graceful handling of date parsing issues

3. Performance Optimization

• Set temperature=0.1 for consistent reports
• Enable memory=True for context retention
• Use planning=True for better task coordination

4. Cost Management

• Monitor API usage in Google Cloud Console
• Set usage alerts for cost control
• Use flash model for development/testing

🔄 Migration from Original Implementation

If migrating from the Llama Stack version:

1. Keep Same Data Source

# Your MCP server continues to work
export USE_MCP_SERVER="true"
export MCP_SERVER_URL="http://localhost:8000"

2. Switch LLM Provider

# From: Llama Stack + OpenAI
# To: CrewAI + Gemini
export LLM_PROVIDER="gemini"
export GEMINI_API_KEY="your-key"

3. Enhanced Capabilities

• ✅ Same JIRA data access via MCP
• ✅ Better multi-agent collaboration
• ✅ More sophisticated reporting
• ✅ Better error handling
• ✅ Standalone deployment (no Llama Stack needed)

🎁 Example Output

The agent generates a professional report like:

# Weekly JIRA Closed Issues Report
**Report Period:** July 15-22, 2025

## Executive Summary
- Total issues closed: 12
- CCITJEN: 5 issues
- CCITRP: 4 issues  
- QEHS: 3 issues

## Detailed Results
[Professional tables with issue details]

🆘 Support

If you encounter issues:

1. Check environment: python crewai_config.py
2. Verify MCP server: Test endpoint directly
3. Check logs: Run with verbose mode
4. Review API limits: Check Gemini quotas

🚀 Next Steps

1. Run basic test: python crewai_gemini_implementation.py
2. Customize prompts: Edit agent backstories and tasks
3. Add more projects: Extend the projects list
4. Schedule reports: Set up cron jobs for automation
5. Add integrations: Connect to Slack, email, etc.

---

🎉 You're ready to use Gemini with CrewAI and your MCP server!

The setup provides the same JIRA data access with enhanced AI capabilities and better architectural design.