MemToolAgent Improves LLM Tool Use via Structured Memory from Past Interactions

MemToolAgent extracts structured memory entries from prior user-agent conversations and dynamically retrieves relevant entries at inference time to improve tool-use accuracy and personalization, without fine-tuning the base LLM. The framework targets the gap between sophisticated dialogue memory systems and tool-using agent memory.