Tool Calling & Agent Loops

Calling one tool once is easy. An agent loop is what happens when Hermes calls a tool, reads the result, decides it needs another tool (or the same one again) to finish the job, and keeps going until the task is actually done — code execution, file I/O, and delegating sub-tasks all included.

Objectives

• Distinguish a single tool call from a multi-step agent loop
• Use --yolo deliberately and understand what it skips (confirmation prompts)
• Run a task that requires Hermes to write, execute, and act on code output in one loop
• Delegate a sub-task to a subagent and inspect what comes back

Lab: a real multi-step agent loop

Give Hermes a task that requires writing a file, then reading it back — two tool calls chained into one loop:

$ hermes -z "Create a file at /tmp/hermes-demo.txt containing some text, then read it back and show me the contents." --cli --yolo

Then push further with code execution as the tool, which forces Hermes to actually run something rather than just narrate an answer:

$ hermes -z "Use Python code execution to compute the 25th Fibonacci number and the first 10 prime numbers. Show your work." --cli --yolo

Finally, delegate a sub-task to a subagent and inspect what it returns:

$ hermes -z "Delegate a subagent to write a haiku about local LLMs, then show me what the subagent returned." --cli --yolo