OpenBanking MCP Server

# OpenBanking MCP Server - Setup Guide ## Quick Start ### 1. Prerequisites Check - [ ] Python 3.8+ installed - [ ] Conda environment `openbanking-backend` exists - [ ] Ollama installed (download from https://ollama.ai) ### 2. Environment Setup ```powershell # Activate conda environment conda activate openbanking-backend # Install Python dependencies pip install PyYAML typing-extensions httpx pydantic ollama # Optional: Install MCP packages (if available) pip install mcp ``` ### 3. Ollama Setup ```bash # Install a model (choose one) ollama pull llama3.2:latest ollama pull llama3.2:3b ollama pull codellama:latest # Start Ollama server (keep this running) ollama serve ``` ### 4. Test the Server ```powershell # Test basic functionality python simple_server.py # Run demo examples python demo.py # Full server (if MCP dependencies are available) python src/main.py ``` ## Server Components ### Core Architecture ``` MCP Host ←→ MCP Server ←→ 4 Specialized Agents ←→ Ollama LLMs ↓ Tool System ↓ Banking APIs/Services ``` ### Agents Configuration 1. **Market Analyst Agent** - Model: llama3.2:latest - Role: Market data analysis and volatility assessment - Tasks: Market trends, technical analysis, economic indicators 2. **Portfolio Manager Agent** - Model: llama3.2:latest - Role: Portfolio optimization and strategy development - Tasks: Asset allocation, rebalancing, performance analysis 3. **Risk Analyst Agent** - Model: llama3.2:latest - Role: Risk assessment and management - Tasks: Risk profiling, stress testing, scenario analysis 4. **Explainability Agent** - Model: llama3.2:latest - Role: Education and explanation - Tasks: SWOT analysis, concept explanation, decision support ### Available Tools #### Portfolio Tools - `analyze_portfolio`: Comprehensive portfolio analysis - `portfolio_optimization`: Asset allocation optimization - `performance_attribution`: Performance breakdown analysis #### Market Tools - `market_analysis`: Current market conditions - `volatility_analysis`: Volatility patterns and forecasting - `sector_analysis`: Sector-specific analysis - `correlation_analysis`: Asset correlation study #### Risk Tools - `assess_risk`: User-specific risk assessment - `simulate_scenarios`: Stress testing and scenarios - `liquidity_risk_analysis`: Liquidity risk evaluation - `tail_risk_analysis`: Extreme event analysis #### Strategy Tools - `recommend_strategy`: Investment strategy recommendations - `tactical_allocation`: Short-term allocation adjustments - `rebalancing_strategy`: Portfolio rebalancing plans - `hedge_strategy`: Risk hedging recommendations #### Analysis Tools - `swot_analysis`: SWOT analysis framework - `explain_concept`: Financial concept explanations - `reverse_simulation`: Reverse engineering analysis - `decision_analysis`: Decision support framework - `trend_analysis`: Trend identification and forecasting ## Configuration Files ### config/config.yaml ```yaml server_name: "openbanking-mcp" development_mode: false ollama: host: "localhost" port: 11434 timeout: 30 agents: - name: "market_analyst" model: "llama3.2:latest" role: "Market Data Analyst" # ... (see full config) tools: - name: "portfolio_analysis" enabled: true ``` ## Troubleshooting ### Common Issues 1. **"Import could not be resolved" errors** - These are IDE warnings, not runtime errors - Code will run correctly if dependencies are installed 2. **Ollama connection failed** ```bash # Check if Ollama is running curl http://localhost:11434/api/tags # Start Ollama if not running ollama serve ``` 3. **Conda environment issues** ```bash # Create environment if missing conda create -n openbanking-backend python=3.11 # Activate environment conda activate openbanking-backend ``` 4. **Dependencies missing** ```bash # Install core dependencies pip install PyYAML httpx pydantic ollama typing-extensions ``` ### Testing Steps 1. **Test Configuration Loading** ```python from src.config.config import Config config = Config.load("config/config.yaml") print(f"Loaded {len(config.agents)} agents") ``` 2. **Test Ollama Connection** ```python import ollama client = ollama.Client() models = client.list() print([m['name'] for m in models['models']]) ``` 3. **Test Agent Communication** ```bash python simple_server.py ``` ## Next Steps 1. **Customize Agents**: Modify agent prompts in `config/config.yaml` 2. **Add Custom Tools**: Create new tool classes in `src/tools/` 3. **Integrate APIs**: Add real market data APIs 4. **Deploy**: Set up production environment 5. **Monitor**: Add logging and monitoring ## File Structure Summary ``` mcpOpenbankingMCPServer/ ├── src/ # Source code │ ├── main.py # Full MCP server │ ├── agents/ # Agent management │ ├── tools/ # Tool implementations │ ├── config/ # Configuration management │ └── utils/ # Utilities ├── config/ # Configuration files ├── simple_server.py # Simplified test server ├── demo.py # Example usage ├── startup.py # Startup script ├── start_server.ps1 # PowerShell script └── requirements.txt # Dependencies ``` ## Support For issues or questions: 1. Check this setup guide 2. Review error messages in terminal 3. Verify Ollama is running and accessible 4. Ensure conda environment is activated 5. Check that all dependencies are installed The server is designed to work even if some components fail, with graceful degradation and helpful error messages.