mcp[patch]: raise a runtime error if the tool call is not available (#318)

mcp-sdk context manager is swallowing some classes of exceptions with no
way to recover the actual error the occurred.

This patch makes a slight improvement to the exception an end user sees
(so it's not an unbound variable).

Unfortunately, we don't have any information about the root cause (we
don't even know whether a timeout can cause this.)

I'll try to recreate this issue:
#314 in
unit tests to see if we can trigger it with timeouts.

Author: eyurtsev

langchain_mcp_adapters/tools.py

@@ -124,6 +124,7 @@ def convert_mcp_tool_to_langchain_tool(
     async def call_tool(
         **arguments: dict[str, Any],
     ) -> tuple[str | list[str], list[NonTextContent] | None]:
+        call_tool_result = None
         if session is None:
             # If a session is not provided, we will create one on the fly
             async with create_session(connection) as tool_session:
@@ -134,6 +135,15 @@ async def call_tool(
                 )
         else:
             call_tool_result = await session.call_tool(tool.name, arguments)
+
+        if call_tool_result is None:
+            raise RuntimeError(
+                "Tool call failed: no result returned from the underlying MCP SDK. "
+                "This may indicate that an exception was handled or suppressed "
+                "by the MCP SDK (e.g., client disconnection, network issue, "
+                "or other execution error)."
+            )
+
         return _convert_call_tool_result(call_tool_result)
 
     meta = tool.meta if hasattr(tool, "meta") else None