From answering to acting

An agent is an LLM given tools (functions/APIs) and a goal. It decides which tool to call, reads the result, and loops until the task is done.

The agent loop

1. Model receives a goal + the tools available.
2. It chooses a tool and arguments.
3. Your code runs the tool and returns the result.
4. The model incorporates it and continues — or finishes.

Examples of tools

• Web search, database queries, sending email.
• Running code, calling internal APIs, updating records.

Building them well

Clear tool descriptions, guardrails on destructive actions, timeouts/retries, and observability. Start with one or two tools before going autonomous.

Takeaway: tools turn a chatbot into a system that gets real work done.