LLM function calls don't scale; code orchestration is simpler, more effective

One common practice for working with MCP tools calls is to put the outputs from a tool back into the LLM as a message, and ask the LLM for the next step. …

Read in full here: