Too many command asking from agent

Since i’ve updated cursor to last version i’m asking way too often by the agent for him to analyse the code by command instead of looking directly into the files