What is an AI agent?
A chatbot is like a friend who answers your questions. An agent is like a friend who can also go do things: look something up, fill in a list, check a calendar, then tell you what they did.
- Anthropic's engineering article "Building effective agents" (December 19, 2024) groups these under "agentic systems" and splits them in two:
- Workflows: the AI and its tools follow steps that you planned in advance.
- Agents: the AI decides its own steps and which tools to use, until the task is done.
- The same article's main advice: start with the simplest solution, and add complexity only when it clearly makes results better. Many tasks need just one well-written prompt.
- Agents fit open-ended problems where you can't predict the number of steps. They cost more, and mistakes can add up, so they need testing and guardrails.
- Current Claude models (Nov 2026): Claude Opus 5.5 (default), Claude Sonnet 5.5, Claude Haiku 5.5, and Claude Fable 5.2 (Anthropic's most capable widely released model).
| Level | What it is | Example |
|---|---|---|
| Single prompt | One question, one answer | "Summarize this email." |
| Workflow | Fixed steps you designed | Summarize, then draft reply, then check tone |
| Agent | AI picks steps and tools | "Find out which orders are late and draft apology emails." |
- Write down 5 tasks you would like to hand off, one per line.
- For each task, ask: "Can one good prompt do it?" If yes, mark it "single prompt".
- If the steps are known and always the same (for example summarize, then draft, then check tone), mark it "workflow".
- If you cannot predict the number of steps or which tools are needed, mark it "agent".
- Look at your list. Most tasks should be single prompts or workflows. Only build an agent when a simpler shape clearly fails.
- Check: every task has a label, and you can say in one sentence why you chose it.
Videos by other creators on YouTube, for reference. They are not made by or affiliated with The Future Market.
Read the video script
[Host] "Everyone says 'AI agents.' Here's what it actually means."
[Text on screen: Chatbot answers. Agent acts.] "A chatbot answers. An agent can take steps: look things up, use tools, then report back."
[Screen: Anthropic's 'Building effective agents' article] "Anthropic's own advice: start with the simplest thing that works."
[Host] "One good prompt first. A workflow next. A full agent only when you really need it."
List 5 tasks you'd love to hand off. For each, write whether it needs a single prompt, a workflow (fixed steps) or an agent (unknown steps). Most will be the first two, and that's good.
ClaudeStuck? Copy a ready-made prompt for Claude
Copy this, open Claude, paste it, and Claude will walk you through the lesson step by step.
For a grown-up: Claude is for adults (18+). Copy this, paste it into Claude, and use the answer to explain the lesson to your child.
I'm learning "Build AI Agents with Claude: No-Code to Code" on The Future Market. Lesson 1: "What is an AI agent?". What the lesson teaches: - Anthropic's engineering article "Building effective agents" (December 19, 2024) groups these under "agentic systems" and splits them in two: - Workflows: the AI and its tools follow steps that you planned in advance. - Agents: the AI decides its own steps and which tools to use, until the task is done. - The same article's main advice: start with the simplest solution, and add complexity only when it clearly makes results better. Many tasks need just one well-written prompt. - Agents fit open-ended problems where you can't predict the number of steps. They cost more, and mistakes can add up, so they need testing and guardrails. - Current Claude models (Nov 2026): Claude Opus 5.5 (default), Claude Sonnet 5.5, Claude Haiku 5.5, and Claude Fable 5.2 (Anthropic's most capable widely released model). - Level - What it is - Example The exercise I'm working on: List 5 tasks you'd love to hand off. For each, write whether it needs a single prompt, a workflow (fixed steps) or an agent (unknown steps). Most will be the first two, and that's good. I'm stuck. Please: 1. Explain the main idea in plain, simple words with one everyday example. 2. Walk me through the exercise one small step at a time. Ask me what I've done so far before giving the next step. 3. Don't just give me the final answer. Help me do it myself, and check my work when I share it.
I'm a parent helping my child (about 8 years old) with a lesson called "What is an AI agent?" from the course "Build AI Agents with Claude: No-Code to Code". Here's what the lesson says: A chatbot is like a friend who answers your questions. An agent is like a friend who can also go do things: look something up, fill in a list, check a calendar, then tell you what they did. The activity: List 5 tasks you'd love to hand off. For each, write whether it needs a single prompt, a workflow (fixed steps) or an agent (unknown steps). Most will be the first two, and that's good. Please explain this so an 8-year-old understands: short sentences, a fun everyday example, and no hard words. Then give me 3 simple questions I can ask my child to check they understood, and a way to do the activity together, one tiny step at a time.
Sources
- Anthropic, Building effective agents: www.anthropic.com/engineering/building-effective-agents
- Anthropic, Models overview: platform.claude.com/docs/en/models/overview
No-code agents in the Claude app
This lesson is part of full access: every course and every level for one payment.
Five patterns that solve most problems
This lesson is part of full access: every course and every level for one payment.
Your first API call (code)
This lesson is part of full access: every course and every level for one payment.
Give Claude a tool
This lesson is part of full access: every course and every level for one payment.
Test it, guard it, then grow it
This lesson is part of full access: every course and every level for one payment.
Final project
This lesson is part of full access: every course and every level for one payment.
Quiz
What's the difference between a workflow and an agent?
A workflow follows steps you planned; an agent decides its own steps and tool use.
What does Anthropic recommend before building an agent?
Start with the simplest solution; add complexity only when it improves results.
What is MCP?
An open standard (announced Nov 25, 2024) for connecting AI apps to data sources and tools.
Name the five workflow patterns.
Prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer.
Where should your API key live?
In the ANTHROPIC_API_KEY environment variable, never in your code.
What does Claude send when it wants to use a tool, and what do you send back?
A tool_use request; you send back a tool_result.
What does the SDK's tool runner do?
Runs your tools when Claude calls them and handles the loop automatically.
What's a safe first rule for agent tools?
Start read-only; require human approval for actions that send, delete or pay.
Final exam
This lesson is part of full access: every course and every level for one payment.
Certificate
This lesson is part of full access: every course and every level for one payment.