Adding custom tools to the managed MCP server

To add custom MCP tools:

1. In your local file system, create a directory structure that is similar to the app/services directory structures in the https://github.com/IBM/data-intelligence-mcp-server repository.
   Create a new directory with at least one subdirectory and 3 files:

   1. An __init__.py file in the new directory
   2. A tools subdirectory
   3. An __init__.py file in the tools subdirectory

   cat << EOF > __init__.py
   EOF
   
   mkdir tools
   
   cat << EOF > tools/__init__.py
   EOF

2. For each new MCP tool, create a new .py file in the tools subdirectory.
   The tool configuration must include the custom tag like in the following example:

   cat << EOF >> ./tools/example_tool.py
   from typing import Any
   
   from app.core.registry import service_registry
   from app.shared.logging import LOGGER
   
   
   @service_registry.tool(
       name="test_tool",
       description="Example custom tool. This demonstrates the structure for custom tools.",
       tags={"custom"}
   )
   async def test_tool(
       resource_id: str,
       action: str = "query"
   ) -> dict[str, Any]:
       """
       Example custom tool 
       Args:
           resource_id: The ID of the resource to operate on
           action: The action to perform (query, update, delete, etc.)
       
       Returns:
           Dictionary containing the operation result
       """
       LOGGER.info(f"Executing custom tool for resource: {resource_id}, action: {action}")
       
       result = {
               "status": "success",
               "resource_id": resource_id,
               "action": action,
               "message": "This is an example custom tool. Replace with actual implementation."
           }
           
       LOGGER.info(f"Custom tool completed successfully for resource: {resource_id}")
       return result
   EOF

3. Synchronize the files with the semantic-agent-service pod and disable tool filtering.
   1. Copy the files:

      POD_NAME=`oc get pods | grep semantic-agent | awk '{print $1}'`
      oc rsync ./ $POD_NAME:/sal/custom-tools

   2. Verify that all files were copied:

      oc rsh deploy/semantic-agent-service  ls -R /sal/custom-tools

   3. Disable tool filtering so that custom tools can be loaded:

      oc set env deploy/semantic-agent-service enable_mcp_tool_filtering=false