AI Agents and Tool Use
How LLMs go beyond chat to plan, use tools, and take actions.
Inspired by learning material from OpenAI
An AI agent is an LLM that doesn't just answer — it plans, uses tools, and takes actions to accomplish goals.
From chatbot to agent
A plain LLM responds to a prompt. An agent can:
- Break a goal into steps
- Call tools (search, calculators, APIs, databases)
- Observe results and decide what to do next
The reasoning loop
Agents often follow a loop: think → act → observe → repeat, until the goal is met. This lets them handle multi-step tasks.
Tool use
By giving the model functions it can call, you extend its abilities far beyond text — fetching live data, running code, or updating records.
Risks and guardrails
Agents that take actions need limits: permissions, validation, and human approval for sensitive steps.
Key takeaways
- Agents plan and act, not just chat.
- Tool use connects LLMs to the real world.
- Guardrails are essential when agents take actions.