I have the same problem.
The Agent (LLM) ends its actions by simulating a tool call in the following format:
Calling My_Tool with input: {“query”:“my query”,“tool”:“my_tool”,“id”:“call_453138186985815f”}
This is especially common when forcing the Agent to call tools sequentially. It will call the first one correctly, but the second one will fail with the aforementioned error.
I think a similar problem was also observed in the following issue.