Checking and Refining the Agent's Reasoning › Combining Verification with Chain-of-Thought

from anthropic import Anthropic

client = Anthropic(api_key=os.getenv("ANTHROPIC_API_KEY"))

problem = """A rectangular garden is 12 meters long and 8 meters wide. 
You want to build a path 1 meter wide around the entire garden. 
What is the area of the path?"""

## Stage 1 & 2: Chain-of-thought solving
response = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": f"""{problem}

Think through this step by step:

1. What are the dimensions of the garden?
2. What will be the dimensions including the path?
3. How can you calculate the path area?

Solve the problem showing your work.""",
        }
    ],
)

solution = response.content[0].text
print("Solution:", solution)

## Stage 3: Verification
verification = client.messages.create(
    model="claude-sonnet-4-5",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": f"""{problem}

Your solution:
{solution}

Verify by:

1. Checking dimensions are correct
2. Confirming area calculations
3. Ensuring you calculated the path area, not the total area

If you find errors, provide a corrected solution.""",
        }
    ],
)

print("\nVerification:", verification.content[0].text)