I’m looking at Grok’s API for an assistant that needs to do more than one thing in a turn—for example, look up an account, check its status, then prepare a response based on the result. How well does Grok handle that kind of multi-step tool calling?

I’m curious whether it reliably makes the next tool call after receiving a result, or whether you have to guide the sequence more tightly. Experiences with retries and tools that return errors would also be useful.