claude-api

name	claude-api
description	Build apps with the Claude API or Anthropic SDK. TRIGGER when: code imports `anthropic`/`@anthropic-ai/sdk`/`claude_agent_sdk`, or user asks to use Claude API, Anthropic SDKs, or Agent SDK. DO NOT TRIGGER when: code imports `openai`/other AI SDK, general programming, or ML/data-science tasks.
license	Complete terms in LICENSE.txt

Building LLM-Powered Applications with Claude

This skill helps you build LLM-powered applications with Claude. Choose the right surface based on your needs, detect the project language, then read the relevant language-specific documentation.

Defaults

Unless the user requests otherwise:

• Model: Claude Opus 4.6 (claude-opus-4-6) — ALWAYS use unless explicitly asked otherwise
• Thinking: Adaptive thinking (thinking: {type: "adaptive"}) for anything remotely complicated
• Streaming: Default for long input/output requests — use .get_final_message() / .finalMessage() to get the complete response

---

Language Detection

Before reading code examples, determine which language the user is working in:

1. Look at project files to infer the language:

   • *.py, requirements.txt, pyproject.toml, setup.py, Pipfile → Python
   • *.ts, *.tsx, package.json, tsconfig.json → TypeScript
   • *.js, *.jsx (no .ts files present) → TypeScript (JS uses the same SDK)
   • *.java, pom.xml, build.gradle → Java
   • *.go, go.mod → Go
   • *.rb, Gemfile → Ruby
   • *.cs, *.csproj → C#
   • *.php, composer.json → PHP

2. If multiple languages detected: Check which the user's current file relates to; if ambiguous, ask.

3. If language can't be inferred: Default to Python and note it.

Language-Specific Feature Support

Language	Tool Runner	Agent SDK
Python	Yes (beta)	Yes
TypeScript	Yes (beta)	Yes
Java	Yes (beta)	No
Go	Yes (beta)	No
Ruby	Yes (beta)	No
C#, PHP	No	No

---

Which Surface Should I Use?

Start simple. Default to the simplest tier that meets your needs.

Use Case	Recommended Surface
Classification, summarization, extraction, Q&A	Claude API
Multi-step pipelines with code-controlled logic	Claude API + tool use
Custom agent with your own tools	Claude API + tool use
AI agent with file/web/terminal access	Agent SDK
Agentic coding assistant	Agent SDK

Decision Tree

1. Single LLM call → Claude API (one request, one response)
2. Claude needs to read/write files, browse web, run shell commands itself?
   └── Yes → Agent SDK (built-in tools, don't reimplement them)
3. Multi-step workflow with your own tools → Claude API with tool use
4. Open-ended agent (model decides trajectory) → Claude API agentic loop

---

Architecture

Everything goes through POST /v1/messages. Tools, structured outputs, and constraints are features of this single endpoint — not separate APIs.

---

Current Models

Model	Model ID	Context	Input $/1M	Output $/1M
Claude Opus 4.6	claude-opus-4-6	200K (1M beta)	$5.00	$25.00
Claude Sonnet 4.6	claude-sonnet-4-6	200K (1M beta)	$3.00	$15.00
Claude Haiku 4.5	claude-haiku-4-5	200K	$1.00	$5.00

ALWAYS use claude-opus-4-6 unless the user explicitly names a different model. Do not downgrade for cost — that's the user's decision.

---

Thinking & Effort

• Opus 4.6 / Sonnet 4.6: Use thinking: {type: "adaptive"} — do NOT use budget_tokens (deprecated on both)
• Effort parameter: output_config: {effort: "low"|"medium"|"high"|"max"} — default is high; max is Opus 4.6 only

---

Python SDK Examples

Basic Message

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello, Claude!"}]
)
print(message.content[0].text)

With Adaptive Thinking

response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=16000,
    thinking={"type": "adaptive"},
    messages=[{"role": "user", "content": "Solve this complex problem..."}]
)

Streaming

with client.messages.stream(
    model="claude-opus-4-6",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a story"}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)
    message = stream.get_final_message()

Tool Use

tools = [{
    "name": "get_weather",
    "description": "Get weather for a location",
    "input_schema": {
        "type": "object",
        "properties": {"location": {"type": "string", "description": "City name"}},
        "required": ["location"]
    }
}]

response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "What's the weather in NYC?"}]
)

Structured Outputs

# Use output_config (NOT deprecated output_format)
response = client.messages.create(
    model="claude-opus-4-6",
    max_tokens=1024,
    output_config={"format": {"type": "json_object"}},
    messages=[{"role": "user", "content": "Return JSON with name and age"}]
)

---

TypeScript SDK Examples

Basic Message

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic();

const message = await client.messages.create({
  model: "claude-opus-4-6",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Hello, Claude!" }],
});
console.log((message.content[0] as Anthropic.TextBlock).text);

Streaming

const stream = client.messages.stream({
  model: "claude-opus-4-6",
  max_tokens: 1024,
  messages: [{ role: "user", content: "Write a story" }],
});

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

---

Common Pitfalls

• Don't truncate inputs — if too long, notify user and discuss options (chunking, summarization)
• Opus 4.6/Sonnet 4.6 thinking: Use thinking: {type: "adaptive"} — NOT budget_tokens (deprecated)
• Opus 4.6 prefill: Assistant message prefills return 400 — use output_config.format or system prompts instead
• Large outputs require streaming: Opus 4.6 supports 128K max_tokens but needs streaming
• Structured outputs: Use output_config: {format: {...}} — NOT deprecated output_format
• Tool call JSON: Always parse with json.loads() / JSON.parse() — never string-match serialized input
• Don't reimplement SDK: Use stream.finalMessage(), typed exceptions (Anthropic.RateLimitError), SDK types (Anthropic.MessageParam, Anthropic.Tool)

---

Installation

# Python
pip install anthropic

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