# MCP Server Test Client
This directory contains a test client for interacting with the Senzing MCP server directly using the MCP Python SDK. It's useful for testing your setup and understanding how the MCP server works.
## Prerequisites
1. **MCP server installed**:
```bash
cd /path/to/senzing-mcp-server
pip install -e .
```
2. **Senzing environment configured**:
```bash
# Required environment variables:
export SENZING_ENGINE_CONFIGURATION_JSON='{...}'
export LD_LIBRARY_PATH=/opt/senzing/lib
export PYTHONPATH=/opt/senzing/sdk/python
```
## Usage
The `senzing_test.py` script provides a unified CLI for testing all 7 MCP tools:
```bash
# List available MCP tools
python senzing_test.py list-tools
# Search for entities by name
python senzing_test.py search "John Smith"
# Get entity details by ID
python senzing_test.py get 1
# Get entity by source record
python senzing_test.py get-record CUSTOMERS 1001
# Find relationship path between entities
python senzing_test.py find-path 1 2
python senzing_test.py find-path 1 2 3 # with max_degrees
# Expand entity network
python senzing_test.py expand 1
python senzing_test.py expand 1 2 # with max_degrees
python senzing_test.py expand 1 2 20 # with max_degrees and max_entities
# Explain why entities are related
python senzing_test.py why 1 2
# Explain how entity was resolved
python senzing_test.py how 1
```
### Built-in Help
Run without arguments to see full usage documentation:
```bash
python senzing_test.py
```
## How It Works
The test client uses the MCP Python SDK to:
1. **Start the MCP server** using `StdioServerParameters`
2. **Connect** via stdin/stdout transport
3. **Initialize** a client session
4. **Call tools** exposed by the MCP server
5. **Parse and display** the JSON results
## Using in Your Own Code
You can use this as a template for your own MCP client applications:
```python
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
# Configure server
server_params = StdioServerParameters(
command="senzing-mcp",
env={
"SENZING_ENGINE_CONFIGURATION_JSON": config_json,
"LD_LIBRARY_PATH": ld_library_path
}
)
# Connect and use
async with stdio_client(server_params) as (read, write):
async with ClientSession(read, write) as session:
await session.initialize()
# Call any MCP tool
result = await session.call_tool("search_entities", {
"attributes": {"NAME_FULL": "John Smith"}
})
# Process result
data = json.loads(result.content[0].text)
```
## Available MCP Tools
The server provides these 7 read-only tools:
1. **search_entities** - Search by name, address, phone, email, etc.
2. **get_entity** - Get entity by ID
3. **get_source_record** - Get entity by source record ID (e.g., CUSTOMERS:1001)
4. **find_path** - Find relationship path between entities
5. **expand_network** - Expand networks of related entities
6. **explain_why_related** - Explain why two entities are related (WHY analysis)
7. **explain_how_resolved** - Explain how entity was resolved (HOW analysis)
## Troubleshooting
**Environment errors:**
```bash
# Ensure environment variables are set
echo $SENZING_ENGINE_CONFIGURATION_JSON
echo $LD_LIBRARY_PATH
```
**Server not found:**
```bash
# Use absolute path if senzing-mcp not in PATH
export SENZING_MCP_COMMAND=/path/to/senzing-mcp-server/venv/bin/senzing-mcp
python senzing_test.py list-tools
```
**Import errors:**
```bash
# Ensure MCP package is installed
pip install mcp
```
## Notes
- This is a **test/example client** - in production, you'd typically use the MCP server through AI assistants (Claude, ChatGPT, Amazon Q)
- The script starts a new MCP server instance for each command
- The server automatically handles SDK initialization
- All commands require proper environment configuration
