Trade Surveillance Support MCP Server

# Trade Surveillance Support MCP Server ## Project Overview This is a Model Context Protocol (MCP) server designed to automate trade surveillance support workflows by: - Parsing user inquiry emails - Searching relevant SQL config files and Java code - Executing Java processes to generate reports - Streamlining user support and investigation ## Implementation Details - **Language**: Python 3.10+ - **Framework**: MCP SDK with FastMCP - **Transport**: STDIO (for local execution) - **Package Manager**: uv (recommended) ## Architecture - **MCP Server** (`trade_surveillance_mcp/server.py`): Provides custom tools to VS Code Copilot - **Email Parser**: Extracts key information from user inquiries - **Config Searcher**: Finds relevant SQL and config files in your repository - **Code Searcher**: Locates Java classes and methods for report generation - **Java Executor**: Runs Java code to generate reports - **Response Generator**: Formats comprehensive summaries ## Available MCP Tools ### 1. parse_email_inquiry Analyzes email content to extract inquiry type, trade IDs, time periods, priority, and suggested actions. ### 2. search_sql_configs (⭐ Metadata-based) Searches through SQL configuration files by **keywords** instead of file paths. Files must have metadata annotations: ```sql -- @keywords: settlement, trade, daily -- @type: compliance_report -- @description: Daily settlement report ``` ### 3. search_java_code (⭐ Metadata-based) Locates Java source files by **keywords** instead of file paths. Classes must have javadoc annotations: ```java /** * @keywords report, generator, settlement * @type report_generator */ ``` ### 4. execute_java_report Executes Java processes with specified config files to generate reports. ### 5. rebuild_metadata_index Rebuilds the search index by scanning all annotated files. Use after adding new files or updating annotations. ### 6. generate_response_summary Creates comprehensive response summaries combining all gathered information. ## Usage with VS Code Copilot 1. **Start a chat**: Open Copilot chat in VS Code 2. **Paste user email**: Copy and paste the support inquiry email 3. **Let Copilot work**: It will automatically: - Parse the email using `parse_email_inquiry` - Search for relevant configs with `search_sql_configs` - Find Java code with `search_java_code` - Execute reports with `execute_java_report` - Generate a summary with `generate_response_summary` Example prompt: ``` I received this email from a trader asking about their transactions from last week: [Email content here] Can you help me investigate and generate the necessary reports? ``` ## Development Guidelines - Follow MCP SDK best practices - Implement async functions for all tools - Use proper logging for debugging - Handle errors gracefully with clear messages - Update search paths to match your repository structure ## Customization Required Before using in production, customize these areas: 1. **Repository Paths**: Update default directories in tool functions 2. **Email Parsing**: Implement NLP or regex patterns for your email format 3. **File Search**: Add actual file searching logic (grep, ripgrep, etc.) 4. **Java Execution**: Configure Java classpath and execution parameters 5. **Config Format**: Adapt to your SQL config file format ## Status ✓ Project setup complete ✓ MCP server implemented with 5 tools ✓ Dependencies installed and tested ✓ VS Code configuration ready **Next Steps**: Customize paths and implement actual search/execution logic for your environment.