I wonder if it's possible for LLM to get confused with the amount of available tools. I think the tool description goes into the context as well. This means we might end up eating up too many tokens. Maybe there should be config options to enable / disable specific tools.