mcp_resume_screening

Job Matching MCP Server

A Model Context Protocol (MCP) server that provides intelligent job matching capabilities. Extract structured job requirements from job descriptions and find/rank candidates from your LlamaCloud resume index.

🚀 Features

Core Job Matching Functions

1. extract_job_requirements - Extract structured data from job description text
2. find_matching_candidates - Find and rank candidates from LlamaCloud index
3. search_candidates_by_skills - Search candidates by specific skills
4. score_candidate_qualifications - Score candidate against job requirements

Additional Functions

• add / subtract / multiply - Basic math functions (backward compatibility)

📋 Available MCP Tools

Job Description Processing

• extract_job_requirements(job_description_text: str) - Extract structured job requirements from text

Candidate Management

• find_matching_candidates(required_qualifications: str, preferred_qualifications: str, top_k: int, enable_reranking: bool) - Find candidates matching job qualifications
• search_candidates_by_skills(skills: str, top_k: int) - Search candidates by specific skills
• score_candidate_qualifications(candidate_resume: str, required_qualifications: str, preferred_qualifications: str, job_title: str, job_description: str) - Score candidate against job requirements

📋 Function Details

1. Extract Job Requirements

extract_job_requirements(jd_text: str) -> str

Input: Job description text (copied from job posting)

Output: JSON string containing:

• title: Job title
• company: Company name
• location: Job location
• required_qualifications: Array of required qualifications
• preferred_qualifications: Array of preferred qualifications
• description: Job summary
• experience_level: Experience level (entry/mid/senior)
• employment_type: Employment type (full-time/contract/etc.)

2. Find and Rank Candidates from LlamaCloud

find_matching_candidates(required_qualifications: str, preferred_qualifications: str, top_k: int, enable_reranking: bool) -> str

Input:

• required_qualifications: Comma-separated required qualifications
• preferred_qualifications: Comma-separated preferred qualifications
• top_k: Maximum candidates to retrieve (default: 10)
• enable_reranking: Whether to enable reranking (default: True)

Output: JSON string containing:

• candidates: Array of candidates with scores and analysis
• total_candidates: Number of candidates found
• search_parameters: Details about search configuration

Key Features:

• Retrieves candidates from LlamaCloud index using semantic search
• Weighted scoring: Required qualifications have higher weight
• Match percentage: Overall compatibility score
• Detailed explanations: For each qualification assessment

3. Search Candidates by Skills

search_candidates_by_skills(skills: str, top_k: int) -> str

Input:

• skills: Comma-separated list of skills or keywords
• top_k: Number of top candidates to retrieve (default: 10)

Output: JSON with matching candidates and their scores

4. Score Candidate Qualifications

score_candidate_qualifications(candidate_resume: str, required_qualifications: str, preferred_qualifications: str, job_title: str, job_description: str) -> str

Input:

• candidate_resume: The candidate's resume text
• required_qualifications: Comma-separated required qualifications
• preferred_qualifications: Comma-separated preferred qualifications
• job_title: Job title for context (optional)
• job_description: Job description for context (optional)

Output: Comprehensive analysis including:

• Strengths and weaknesses
• Detailed scoring breakdown
• Hiring recommendations
• Role fit assessment

🛠️ Setup & Configuration

1. Configure API Keys & Settings

You have two options for configuration:

Option A: Environment Variables (Recommended for Production)

# Required: OpenAI API Key
export OPENAI_API_KEY="your-openai-api-key"

# Required: LlamaCloud Configuration  
export LLAMA_CLOUD_API_KEY="your-llamacloud-api-key"
export LLAMA_CLOUD_INDEX_NAME="your-resume-index-name"
export LLAMA_CLOUD_PROJECT_NAME="your-project-name"
export LLAMA_CLOUD_ORGANIZATION_ID="your-organization-id"

# Optional: Server Configuration
export PORT="8080"
export HOST="0.0.0.0"
export REQUEST_TIMEOUT="30.0"
export OPENAI_TEMPERATURE="0.1"

Option B: Direct Configuration in config.py (For Local Development)

1. Open config.py and replace the placeholder values:

   # Replace these placeholder values with your actual API keys:
   OPENAI_API_KEY = "your-actual-openai-api-key-here"
   LLAMA_CLOUD_API_KEY = "your-actual-llamacloud-api-key-here"
   LLAMA_CLOUD_ORGANIZATION_ID = "your-actual-org-id-here"
   LLAMA_CLOUD_INDEX_NAME = "your-actual-index-name"
   

2. ⚠️ Security Warning: If you edit config.py directly, never commit your API keys to version control!

Getting Your API Keys:

• OpenAI API Key: Get from OpenAI Platform
• LlamaCloud API Key: Get from LlamaCloud Console
• LlamaCloud Org ID: Found in your LlamaCloud project settings
• LlamaCloud Index Name: The name of your resume index in LlamaCloud

Without LlamaCloud credentials, the server uses mock candidate data for testing.

2. Protect Your API Keys (Important!)

If you're planning to commit this code to version control, create a .gitignore file to protect your sensitive information:

# Create .gitignore file
cat > .gitignore << EOF
# Environment variables and sensitive files
.env
.env.local
.env.production
config_local.py

# Python
__pycache__/
*.pyc
*.pyo
*.pyd
.Python
.venv/
venv/

# IDE
.vscode/
.idea/
*.swp
*.swo

# OS
.DS_Store
Thumbs.db
EOF

Alternative: You can also create a separate config_local.py file with your actual keys and import it in config.py, then add config_local.py to .gitignore.

3. Install Dependencies

# Install using uv (recommended)
uv install

# Or using pip
pip install fastmcp httpx

4. Run the Server

python server.py

Server starts on http://localhost:8080/mcp (or PORT environment variable)

🧪 Testing

Run the comprehensive test suite:

# Start the server first
python server.py

# In another terminal, run tests
python test_server.py

The test suite will:

• ✅ Extract job requirements from sample JD
• ✅ Retrieve and rank candidates from LlamaCloud (or mock data)
• ✅ Perform detailed analysis of top candidate
• ✅ Test backward compatibility functions

📊 How It Works

Architecture Overview

1. Job Description Processing

   • Uses OpenAI to extract structured requirements from free-form JD text
   • Separates required vs. preferred qualifications
   • Extracts metadata (title, company, location, etc.)

2. Candidate Retrieval

   • Queries LlamaCloud index using semantic search
   • Builds search query from job requirements
   • Retrieves top candidates with similarity scores

3. Intelligent Scoring

   • Uses OpenAI to score each candidate (0-2 scale)
   • 0: Not Met, 1: Somewhat Met, 2: Strongly Met
   • Required qualifications weighted 2x in final score
   • Provides explanations for each score

4. Match Calculation

   Weighted Score = (Required Total × 2) + Preferred Total
   Match % = (Weighted Score / Max Possible Score) × 100
   

Data Flow

graph TD
    A[Job Description Text] --> B[extract_job_requirements]
    B --> C[Structured Requirements JSON]
    C --> D[find_matching_candidates]
    E[LlamaCloud Index] --> D
    D --> F[Ranked Candidates with Scores]
    C --> G[score_candidate_qualifications]
    H[Individual Resume] --> G
    G --> I[Detailed Analysis & Recommendations]

🎯 Use Cases

1. Automated Resume Screening

# Extract requirements from job posting
job_reqs = extract_job_requirements(job_posting_text)

# Find top candidates from your resume database
top_candidates = find_matching_candidates("Python, JavaScript, React", "AWS, Docker", 10, True)

# Get detailed analysis of promising candidates
for candidate in top_5:
    analysis = score_candidate_qualifications(candidate['resume'], job_reqs, candidate['name'])

2. Hiring Pipeline Integration

• ATS Integration: Automatically score incoming applications
• Recruiter Tools: Provide data-driven candidate rankings
• Interview Prep: Generate candidate-specific interview questions

3. Job Market Analysis

• Requirement Trends: Track common qualifications across postings
• Candidate Gap Analysis: Identify missing skills in candidate pool
• Salary Benchmarking: Correlate requirements with compensation data

🚀 Deployment

Prerequisites

Make sure you have the following set up:

• Python 3.10+
• UV package manager
• Google Cloud SDK (gcloud)
• Project ID configured: export PROJECT_ID=<your-project-id>
• Artifact Registry repository created: remote-mcp-servers

Local Development

# 1. Install dependencies
uv install

# 2. Configure API keys (choose one method):

# Method A: Set environment variables
export OPENAI_API_KEY="your-openai-api-key"
export LLAMA_CLOUD_API_KEY="your-llamacloud-api-key"
export LLAMA_CLOUD_INDEX_NAME="your-index-name"
export LLAMA_CLOUD_ORGANIZATION_ID="your-org-id"

# Method B: Edit config.py directly (see configuration section above)

# 3. Run server
python server.py

The server will start on http://localhost:8080/mcp and log which configuration it's using:

[INFO]: LlamaCloudService initialized with index: your-index-name
[INFO]: MCP server starting on 0.0.0.0:8080

Quick Configuration Test:

You can verify your configuration is working by running:

python -c "from config import OPENAI_API_KEY, LLAMA_CLOUD_API_KEY, LLAMA_CLOUD_INDEX_NAME; print(f'OpenAI: {OPENAI_API_KEY[:10]}..., LlamaCloud: {LLAMA_CLOUD_API_KEY[:10]}..., Index: {LLAMA_CLOUD_INDEX_NAME}')"

If you see placeholder values like "your-openai-api-key-here", your configuration needs to be updated.

Docker Deployment

# Build image
docker build -t job-matching-mcp .

# Run container
docker run -p 8080:8080 \
  -e OPENAI_API_KEY="your-key" \
  -e LLAMA_CLOUD_API_KEY="your-key" \
  job-matching-mcp

Google Cloud Run Deployment

📚 Reference Documentation: Build and Deploy a Remote MCP Server to Google Cloud Run in Under 10 Minutes

Initial Deployment

# Build and push to Artifact Registry
gcloud builds submit --region=us-central1 \
  --tag us-central1-docker.pkg.dev/$PROJECT_ID/remote-mcp-servers/mcp-server:latest

# Deploy to Cloud Run
gcloud run deploy mcp-server \
  --image us-central1-docker.pkg.dev/$PROJECT_ID/remote-mcp-servers/mcp-server:latest \
  --region=us-central1 \
  --no-allow-unauthenticated \
  --set-env-vars OPENAI_API_KEY="your-key",LLAMA_CLOUD_API_KEY="your-key"

Redeployment Steps

After making code changes to your MCP server, follow these steps to redeploy:

Step 1: Rebuild the container and push to Artifact Registry

gcloud builds submit --region=us-central1 \
  --tag us-central1-docker.pkg.dev/$PROJECT_ID/remote-mcp-servers/mcp-server:latest

Step 2: Re-deploy the updated container to Cloud Run

gcloud run deploy mcp-server \
  --image us-central1-docker.pkg.dev/$PROJECT_ID/remote-mcp-servers/mcp-server:latest \
  --region=us-central1 \
  --no-allow-unauthenticated

Step 3: Test the deployment (optional)

Start the Cloud Run proxy to test your updated server:

gcloud run services proxy mcp-server --region=us-central1

Then run your test script:

uv run test_server.py

📁 Project Structure

mcp-on-cloudrun/
├── config.py                 # Configuration constants
├── models.py                 # Data structures
├── server.py                 # Main MCP server
├── Dockerfile               # Container configuration
├── pyproject.toml           # Python dependencies
├── services/
│   ├── openai_service.py    # OpenAI API integration
│   └── llamacloud_service.py # LlamaCloud integration
├── tools/
│   ├── math_tools.py        # Math operations (add, subtract, multiply)
│   ├── job_tools.py         # Job description extraction
│   └── candidate_tools.py   # Candidate search and scoring
└── test_server.py           # Test client

🔧 Customization

Scoring Criteria

Modify the scoring prompts in services/openai_service.py to adjust evaluation criteria:

• Change scoring scale (0-2 to 0-5, etc.)
• Adjust weighting between required/preferred qualifications
• Add domain-specific evaluation criteria

LlamaCloud Integration

For production deployment with real candidate data:

1. Set up LlamaCloud account and create resume index
2. Configure environment variables in config.py
3. Replace mock candidate data with actual LlamaCloud API calls

OpenAI Model Selection

Change the model in config.py:

OPENAI_MODEL = "gpt-4o-mini"  # Fast and cost-effective
# OPENAI_MODEL = "gpt-4o"     # Higher quality, more expensive

📈 Performance & Scaling

• Concurrent Requests: FastMCP handles multiple simultaneous job matching requests
• Caching: Consider implementing Redis for frequent job requirement extractions
• Rate Limiting: OpenAI API has rate limits; implement queuing for high-volume usage
• Cost Optimization: Use gpt-4o-mini for most operations, gpt-4o for critical analysis

🔐 Security Considerations

Security Notes

• Always use --no-allow-unauthenticated to require authentication for Cloud Run
• Ensure users have the roles/run.invoker IAM role to access the server
• Use the Cloud Run proxy for local testing with authentication
• API Keys: Never commit API keys to version control
• Input Validation: Server validates all inputs and handles malformed data
• Error Handling: Graceful degradation when external services are unavailable
• Data Privacy: Resume data processed through OpenAI; consider data retention policies

🔍 Troubleshooting

Common Issues:

Configuration Issues:

1. "Invalid API key" errors:

   • Check that your API keys are correctly set in config.py or environment variables
   • Verify API keys are valid and have proper permissions
   • For OpenAI: Ensure you have credits/billing set up

2. "LlamaCloud index not found":

   • Verify LLAMA_CLOUD_INDEX_NAME matches your actual index name
   • Check LLAMA_CLOUD_ORGANIZATION_ID is correct
   • Ensure your LlamaCloud API key has access to the specified index

3. Server shows placeholder values:

   • If you see "your-openai-api-key-here" in logs, your config isn't loading properly
   • Check that you've either set environment variables OR edited config.py directly
   • Restart the server after making configuration changes

Deployment Issues:

4. Authentication errors: Ensure Cloud Run proxy is running and you have proper IAM roles
5. Build failures: Check Dockerfile and dependencies in pyproject.toml
6. Port conflicts: Use lsof -ti:8080 | xargs kill -9 to free up port 8080

Logs:

View Cloud Run logs:

gcloud run services logs tail mcp-server --region=us-central1

View local server logs:

# Server logs are printed to console when running locally
python server.py

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Add tests for new functionality
4. Ensure all tests pass
5. Submit a pull request

📄 License

MIT License - see LICENSE file for details.

Ready to revolutionize your hiring process with AI-powered job matching! 🎯✨