Skip to main content
Glama
vic3custodio

Trade Surveillance Support MCP Server

by vic3custodio
README.md
# Trade Surveillance Support MCP Server

An MCP (Model Context Protocol) server designed to automate trade surveillance support workflows by integrating with your existing SQL configs and Java code repositories.

## Overview

This MCP server enables you to:
- **Parse user inquiry emails** - Extract key information from support emails automatically
- **Search SQL config files** - Find relevant database queries and configurations
- **Search Java code** - Locate report generation classes and methods
- **Execute Java reports** - Run Java processes to generate data and reports
- **Generate response summaries** - Create comprehensive responses for user inquiries

## Installation

### Prerequisites
- Python 3.10 or higher
- `uv` package manager (recommended) or `pip`

### Install with uv (recommended)

```bash
# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install the MCP server
uv pip install -e .
```

### Install with pip

```bash
pip install -e .
```

## Configuration

### Setting up with VS Code

1. Open VS Code Settings
2. Search for "MCP"
3. Add a new MCP server configuration:

```json
{
  "mcp.servers": {
    "trade-surveillance": {
      "type": "stdio",
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/mcp_test_2",
        "run",
        "trade-surveillance-mcp"
      ]
    }
  }
}
```

### Setting up with Claude Desktop

Add to your `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "trade-surveillance": {
      "command": "uv",
      "args": [
        "--directory",
        "/absolute/path/to/mcp_test_2",
        "run",
        "trade-surveillance-mcp"
      ]
    }
  }
}
```

**macOS/Linux:** `~/Library/Application Support/Claude/claude_desktop_config.json`  
**Windows:** `%APPDATA%\Claude\claude_desktop_config.json`

## šŸŽÆ Keyword-Based Search (No File Paths!)

Instead of searching by file paths, this MCP server uses **metadata annotations** so Copilot can find files by what they do:

**Example SQL annotation:**
```sql
-- @keywords: trade, settlement, daily, reconciliation
-- @type: compliance_report
-- @description: Daily trade settlement reconciliation report
```

**Example Java annotation:**
```java
/**
 * @keywords settlement, report, generator
 * @type report_generator
 * @description Generates daily settlement reports
 */
```

**Result:** Copilot searches by keywords like "settlement report" instead of file paths! 

šŸ“š **Documentation:**
- [QUICKSTART.md](QUICKSTART.md) - Get started in 5 minutes
- [METADATA_GUIDE.md](METADATA_GUIDE.md) - Complete annotation guide
- `examples/` - Annotated SQL and Java examples

## Usage

### With GitHub Copilot in VS Code

1. Open a chat with Copilot
2. Paste a user inquiry email
3. Copilot will automatically use the MCP tools to:
   - Parse the email
   - Search for relevant configs **by keywords** (not file paths!)
   - Search for Java code **by keywords**
   - Execute reports
   - Generate a response

Example prompt:
```
I received this email from a user:

[Paste email content here]

Can you help me investigate and generate the necessary reports?
```

### Available Tools

#### 1. `parse_email_inquiry`
Extracts key information from user inquiry emails including inquiry type, trade IDs, time periods, and priority.

#### 2. `search_sql_configs` ⭐ **Metadata-based search**
Searches through your SQL configuration files by **keywords** instead of file paths. Files are searched using metadata annotations (see METADATA_GUIDE.md).

#### 3. `search_java_code` ⭐ **Metadata-based search**
Locates Java classes and methods by **keywords** instead of file paths. Classes are found using javadoc annotations (see METADATA_GUIDE.md).

#### 4. `execute_java_report`
Runs Java processes with the appropriate config files to generate reports.

#### 5. `rebuild_metadata_index`
Rebuilds the search index by scanning all annotated SQL configs and Java files. Run this after adding new files or updating annotations.

#### 6. `generate_response_summary`
Creates a comprehensive summary response for the user.

## Project Structure

```
mcp_test_2/
ā”œā”€ā”€ trade_surveillance_mcp/
│   ā”œā”€ā”€ __init__.py
│   └── server.py          # Main MCP server implementation
ā”œā”€ā”€ pyproject.toml          # Project dependencies
ā”œā”€ā”€ README.md
└── .github/
    └── copilot-instructions.md
```

## Development

### Running Locally

```bash
# Run the server directly
uv run trade-surveillance-mcp

# Or with Python
python -m trade_surveillance_mcp.server
```

### Customization

You'll need to customize the server to work with your specific repository structure:

1. **Update search paths** - Modify `config_directory` and `code_directory` parameters
2. **Implement email parsing** - Add your email parsing logic in `parse_email_inquiry`
3. **Add file search** - Implement actual file searching in `search_sql_configs` and `search_java_code`
4. **Configure Java execution** - Add your Java classpath and execution logic in `execute_java_report`

### Connecting to Your Repository

Point the MCP server to your actual config and code repositories:

```python
# Example: Update default directories
@mcp.tool()
async def search_sql_configs(
    search_term: str,
    config_directory: str = "/path/to/your/sql/configs"
):
    # Your implementation
```

## Next Steps

1. āœ… **MCP server is ready!** - Restart VS Code to load it
2. šŸ“ **Annotate your files** - Add metadata keywords to your SQL configs and Java code ([QUICKSTART.md](QUICKSTART.md))
3. šŸ” **Build the index** - Use `rebuild_metadata_index` tool in Copilot
4. āš™ļø **Customize paths** - Update default directories in `server.py` to your repos
5. šŸŽÆ **Test with Copilot** - Paste a user email and let Copilot search by keywords!

### Key Innovation: No More File Path Search! šŸŽ‰

Your SQL configs and Java files are now **searchable by keywords**:
- Copilot finds files by what they do, not where they are
- Search "settlement report" instead of remembering `configs/reports/daily/settlement_v2.sql`
- See examples in `examples/configs/` and `examples/src/`

## Troubleshooting

### Server not appearing in VS Code
- Check the MCP server logs in VS Code Output panel
- Verify the absolute path in configuration is correct
- Ensure `uv` is installed and in PATH

### Python version issues
- Ensure Python 3.10+ is installed: `python --version`
- Use `uv` for better environment management

### Java execution errors
- Verify Java is installed: `java -version`
- Check classpath configuration
- Ensure config files are accessible

## Resources

- [MCP Documentation](https://modelcontextprotocol.io)
- [FastMCP Documentation](https://github.com/modelcontextprotocol/python-sdk)
- [VS Code MCP Integration](https://code.visualstudio.com/docs/copilot/chat/mcp)

## License

MIT

TDQS

A4/5.0

Scored across 6 tools

Disambiguation5/5

Each tool has a clearly distinct purpose with no overlap. The tools cover different stages of a workflow: parse_email_inquiry for input analysis, search_sql_configs/search_java_code for resource discovery, execute_java_report for execution, generate_response_summary for output formatting, and rebuild_metadata_index for maintenance. An agent can easily distinguish between them based on their specific functions.

Naming Consistency5/5

All tools follow a consistent verb_noun pattern with snake_case naming. The verbs are clear and descriptive (execute, generate, parse, rebuild, search), and the nouns precisely indicate the target resource or action. There are no deviations in naming conventions across the six tools.

Tool Count5/5

Six tools is well-scoped for a trade surveillance support server. Each tool serves a distinct purpose in the workflow, from inquiry parsing to report generation and response formatting. The count is neither too thin nor bloated, providing comprehensive coverage without redundancy.

Completeness4/5

The toolset covers the core trade surveillance workflow effectively: parsing inquiries, searching for relevant configurations and code, executing reports, and generating responses. A minor gap exists in direct data manipulation or validation tools (e.g., validate_trade_data, update_config), but agents can work around this using the existing search and execution tools. The surface supports end-to-end processing without dead ends.

Maintenance

ActivityInactive
ResponsivenessNo issues