Checked other resources

• I added a very descriptive title to this issue.
• I used the GitHub search to find a similar issue and didn't find any.
• I searched the LangChain documentation for a solution.

Example Code

import asyncio
from langchain_core.tools import StructuredTool
from pydantic import BaseModel

class MyInput(BaseModel):
    path: str

async def my_coro(path: str):
    return f"content of {path}"

tool = StructuredTool(
    name="read_file",
    description="Read a file",
    args_schema=MyInput,
    coroutine=my_coro,
)

# This works
async def main():
    result = await tool.ainvoke({"path": "/tmp/test.txt"})
    print(f"async: {result}")

asyncio.run(main())

# This fails
result = tool.invoke({"path": "/tmp/test.txt"})
# NotImplementedError: StructuredTool does not support sync invocation.

Error Message and Stack Trace (if applicable)

NotImplementedError: StructuredTool does not support sync invocation.

Description

StructuredTool._run() only checks self.func. When func is None but coroutine is available (like MCP tools from langchain-mcp-adapters), it raises NotImplementedError instead of falling back to async execution.

The contrast with ainvoke() is notable: ainvoke() correctly falls back to run_in_executor() when coroutine is None. But _run() does not have the symmetric fallback.

This prevents MCP tools (which are async-only) from working with frameworks that use sync invoke() (e.g., LangGraph's ToolNode._execute_tool_sync).

System Info

• langchain-core: 0.3.x
• Python: 3.11+
• OS: macOS