Build It: Terminal Quiz Bot Powered by Claude

What we're building

A terminal script that asks for a topic, calls Claude to get 5 multiple-choice questions, asks them one at a time, tracks your score, and shows the final result. Real code, ~80 lines.

# pyproject.toml: anthropic, pydantic import json from pydantic import BaseModel, Field from anthropic import Anthropic class Question(BaseModel): prompt: str options: list[str] = Field(min_length=4, max_length=4) correct_index: int = Field(ge=0, le=3) explanation: str class QuizSet(BaseModel): topic: str questions: list[Question]Pydantic models define the exact shape we expect from the LLM.

client = Anthropic() def generate_quiz(topic: str, n: int = 5) -> QuizSet: prompt = f"""Generate exactly {n} multiple-choice questions about: {topic} Return ONLY valid JSON, no markdown fence, matching this shape: {{ "topic": "", "questions": [ {{"prompt": "", "options": ["a","b","c","d"], "correct_index": 0, "explanation": ""}} ] }} Target middle-school difficulty. No trick questions.""" response = client.messages.create( model="claude-opus-4-7", max_tokens=2000, messages=[{"role": "user", "content": prompt}], ) raw = response.content[0].text.strip() if raw.startswith("```"): raw = raw.strip("`").split("\n", 1)[1].rsplit("\n", 1)[0] return QuizSet.model_validate_json(raw)One LLM call returns the whole quiz. Pydantic will reject malformed output.

def run_quiz(quiz: QuizSet) -> int: score = 0 for i, q in enumerate(quiz.questions, start=1): print(f"\n--- Question {i}/{len(quiz.questions)} ---") print(q.prompt) for idx, opt in enumerate(q.options): print(f" {idx+1}. {opt}") while True: raw = input("Your answer (1-4): ").strip() if raw in {"1","2","3","4"}: break print("Please enter 1, 2, 3, or 4.") chosen = int(raw) - 1 if chosen == q.correct_index: print("Correct!") score += 1 else: correct = q.options[q.correct_index] print(f"Nope — answer was: {correct}") print(f"Why: {q.explanation}") return score def main(): topic = input("Topic? ").strip() or "the solar system" print(f"Generating quiz about {topic}") try: quiz = generate_quiz(topic) except Exception as e: print(f"Quiz generation failed: {e}") return final = run_quiz(quiz) print(f"\nFinal score: {final}/{len(quiz.questions)}") if __name__ == "__main__": main()The main loop: ask, validate input, score, report.

Mini-exercise

1. Run the quiz on three different topics
2. Add a --difficulty flag (easy/medium/hard) and wire it into the prompt
3. After the quiz, ask Claude to explain one question in simpler terms
4. Save results to quiz_history.json

Big idea: an LLM + a typed schema + a simple loop is a stunningly powerful base for any interactive tool. You've now built the skeleton every AI tutor app shares.