Question

You are building a conversational agent that lets network engineers ask natural-language questions like 'Is GigabitEthernet0/1 up on core-rtr-1?'. The agent uses an LLM connected to an MCP server exposing a get_interface_status(device, interface) tool that queries devices via RESTCONF. During testing, when a user asks about interface status, the LLM sometimes replies with a confident, fabricated 'up/up' answer instead of invoking the tool. Which change best ensures the agent returns accurate, live network data rather than hallucinated results?

Accepted Answer

Configure the agent to require tool invocation for factual device queries and instruct the LLM (via system prompt and tool schema) never to answer interface state from its own knowledge

Answer

Increase the LLM's temperature setting so it produces more varied responses and is less likely to repeat a canned 'up/up' answer

Answer

Add more example interface outputs into the system prompt so the model has richer training context to infer the correct state

Answer

Switch the MCP server transport from stdio to HTTP so the LLM can reach the tool faster and be less tempted to answer directly