Claude Api

Claude API Skill (Anthropic SDK)

Step 1: Detect Language and Install SDK

Scan the project for package.json, requirements.txt, pyproject.toml, build.gradle, or go.mod to identify the language.

# Python
pip install anthropic

# Node.js / TypeScript
npm install @anthropic-ai/sdk

# Go
go get github.com/anthropics/anthropic-sdk-go

---

Step 2: Choose the Right Model

| Model | Best For | Speed | Cost | |-------|---------|-------|------| | claude-opus-4-5 | Complex reasoning, coding, research | Slower | Higher | | claude-sonnet-4-5 | Balanced - most tasks | Medium | Medium | | claude-haiku-3-5 | Fast, lightweight tasks, classification | Fast | Lower |

Default to claude-opus-4-5 for new implementations unless the user specifies otherwise or speed/cost is a constraint.

---

Step 3: Basic API Call

Python

import anthropic
import os

client = anthropic.Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])

message = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Explain the CAP theorem in simple terms."}
    ]
)
print(message.content[0].text)

TypeScript / Node.js

import Anthropic from '@anthropic-ai/sdk'

const client = new Anthropic()  // reads ANTHROPIC_API_KEY from env

const message = await client.messages.create({
  model: 'claude-opus-4-5',
  max_tokens: 1024,
  messages: [
    { role: 'user', content: 'Explain the CAP theorem in simple terms.' }
  ],
})
console.log(message.content[0].text)

---

Step 4: System Prompts and Multi-Turn Conversations

conversation_history = []

def chat(user_message: str) -> str:
    conversation_history.append({"role": "user", "content": user_message})

    response = client.messages.create(
        model="claude-opus-4-5",
        max_tokens=2048,
        system="You are a senior software engineer. Be concise and precise.",
        messages=conversation_history
    )

    assistant_message = response.content[0].text
    conversation_history.append({"role": "assistant", "content": assistant_message})
    return assistant_message

---

Step 5: Streaming

Use streaming for long outputs to reduce perceived latency:

# Python - streaming
with client.messages.stream(
    model="claude-opus-4-5",
    max_tokens=4096,
    messages=[{"role": "user", "content": "Write a detailed explanation of neural networks"}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

// TypeScript - streaming
const stream = client.messages.stream({
  model: 'claude-opus-4-5',
  max_tokens: 4096,
  messages: [{ role: 'user', content: 'Write a detailed explanation of neural networks' }],
})

for await (const chunk of stream) {
  if (chunk.type === 'content_block_delta' && chunk.delta.type === 'text_delta') {
    process.stdout.write(chunk.delta.text)
  }
}

---

Step 6: Tool Use (Function Calling)

tools = [
    {
        "name": "get_weather",
        "description": "Get current weather for a location",
        "input_schema": {
            "type": "object",
            "properties": {
                "location": {"type": "string", "description": "City name, e.g. 'London'"},
                "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
            },
            "required": ["location"],
        },
    }
]

def get_weather(location: str, unit: str = "celsius") -> dict:
    # Your actual implementation
    return {"temperature": 15, "condition": "cloudy", "location": location}

# Agentic loop
messages = [{"role": "user", "content": "What's the weather in Tokyo?"}]

while True:
    response = client.messages.create(
        model="claude-opus-4-5",
        max_tokens=1024,
        tools=tools,
        messages=messages
    )

    messages.append({"role": "assistant", "content": response.content})

    if response.stop_reason == "end_turn":
        # Final text response - done
        for block in response.content:
            if hasattr(block, "text"):
                print(block.text)
        break

    # Process tool calls
    tool_results = []
    for block in response.content:
        if block.type == "tool_use":
            if block.name == "get_weather":
                result = get_weather(**block.input)
                tool_results.append({
                    "type": "tool_result",
                    "tool_use_id": block.id,
                    "content": str(result),
                })

    messages.append({"role": "user", "content": tool_results})

---

Step 7: Vision (Multimodal)

import base64

# From URL
response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[{
        "role": "user",
        "content": [
            {
                "type": "image",
                "source": {"type": "url", "url": "https://example.com/chart.png"},
            },
            {"type": "text", "text": "What does this chart show?"},
        ],
    }]
)

# From file (base64)
with open("diagram.png", "rb") as f:
    image_data = base64.standard_b64encode(f.read()).decode("utf-8")

response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[{
        "role": "user",
        "content": [
            {
                "type": "image",
                "source": {"type": "base64", "media_type": "image/png", "data": image_data},
            },
            {"type": "text", "text": "Describe this diagram."},
        ],
    }]
)

---

Step 8: Prompt Caching (Cost Optimization)

Cache large system prompts or repeated context to reduce costs by up to 90%:

# Mark cacheable content with cache_control
response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    system=[
        {
            "type": "text",
            "text": "You are an expert in our codebase. " + very_long_codebase_context,
            "cache_control": {"type": "ephemeral"}  # cache this block
        }
    ],
    messages=[{"role": "user", "content": user_question}]
)

# Check cache usage in response
print(response.usage.cache_creation_input_tokens)  # tokens written to cache
print(response.usage.cache_read_input_tokens)       # tokens read from cache

Caching rules:

• Cache TTL: 5 minutes (refreshed on each cache hit)
• Minimum cacheable block: 1024 tokens (Haiku: 2048)
• Mark the largest, most stable prefix with cache_control

---

Step 9: Extended Thinking

For complex reasoning tasks (math, coding, analysis):

response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=16000,
    thinking={
        "type": "enabled",
        "budget_tokens": 10000  # allow up to 10k tokens of internal thinking
    },
    messages=[{"role": "user", "content": "Prove that there are infinitely many primes."}]
)

for block in response.content:
    if block.type == "thinking":
        print("Thinking:", block.thinking[:200], "...")  # internal reasoning
    elif block.type == "text":
        print("Answer:", block.text)

---

Step 10: Batch Processing

For large-scale offline processing (50% cost reduction):

# Create batch
batch = client.messages.batches.create(requests=[
    {
        "custom_id": f"task-{i}",
        "params": {
            "model": "claude-haiku-3-5",
            "max_tokens": 512,
            "messages": [{"role": "user", "content": text}]
        }
    }
    for i, text in enumerate(texts)
])

# Poll until complete
import time
while True:
    batch = client.messages.batches.retrieve(batch.id)
    if batch.processing_status == "ended":
        break
    time.sleep(60)

# Retrieve results
for result in client.messages.batches.results(batch.id):
    print(result.custom_id, result.result.message.content[0].text)

---

Application Tiers

| Tier | Pattern | When to Use | |------|---------|-------------| | 1 | Single API call | Classification, extraction, summarization | | 2 | Multi-turn + tools | Research assistant, coding helper | | 3 | Managed Agents | Long-running autonomous tasks |

---

Common Mistakes

• Hardcoding API keys - always use ANTHROPIC_API_KEY environment variable
• Not streaming long outputs - add streaming to reduce latency for responses > 500 tokens
• Ignoring stop_reason - check response.stop_reason; "tool_use" means loop continues
• Skipping prompt caching - add cache_control to large system prompts to cut costs
• Using non-serializable tool results - always stringify tool results before returning
• Forgetting to append assistant + tool_result to messages - the agentic loop requires both

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Jignesh-Ponamwar
• Source: Jignesh-Ponamwar/skills-mcp
• License: Apache-2.0
• Homepage: https://skills-mcp-jignesh.vercel.app/

Install and usage instructions live in the source repository linked above.