Docs → chatbots → agents
Build-along · ~8 min at your own keyboard
A document tells. A chatbot answers. An agent acts — it uses tools to do something, not just reply. That's a starting point, not a full definition: agents vary on separate dials — what tools they can use, whether they run on a schedule or only when asked, whether they remember between sessions, and whether a person approves before anything leaves. You'll meet those dials later. For now the question for every task is: does this need telling, answering, or doing? Two more distinctions you'll need. The model is the brain; the harness is the program wrapped around it that gives it tools and memory. Claude Code is a harness, and the same model behaves differently in a different one. And a workflow is not an agent: in a workflow you fix the steps in advance and the model fills in each one; in an agent the model decides the next step itself. Workflows are more predictable. Use an agent only when the steps can't be known ahead of time.
The long version: Documents, chatbots and agents ↗
Do this
0/3Checkpoints
- ▸You should now see your six items each carrying a label: document, chatbot, or agent.
- ▸You should now see at least two chatbot items marked as copy-paste work.
- ▸You should now see one sentence with a trigger, an action and a destination.
The three ways this goes wrong
- Every item ends up labeled chatbot. — Ask of each one: do I copy the output somewhere afterwards? If yes, it is an agent candidate.
- Your agent sentence has no trigger ('I could use AI to help with invoices'). — Rewrite it as: when X happens, do Y, save it to Z.
- You list tools instead of tasks (ChatGPT, Claude). — List the jobs the tools do for you, not the tool names.
Where these ideas come from: Model, Harness, Agent (AI Hero dictionary). Paraphrased here.
Share what happened
Paste your labeled list and the agent-version sentence.
Strip API keys, passwords, and tokens before you paste. Swap them for [KEY].
🔒 Don't paste confidential, personal-health, or payment info.