Chapter 18: Claude & the Anthropic API

Last Updated: 2026-03

18.2 Authentication

The Anthropic API key is stored:

• n8n: In the n8n credential vault (never in workflow code)
• WordPress/PHP: In WordPress options, encrypted using ITI utilities
• Swift: In the macOS Keychain via KeychainService
• Python: In .env as ANTHROPIC_API_KEY, loaded via python-dotenv
• Rust/Tauri: Via the Tauri Keychain plugin, loaded into IPC commands at runtime

Warning: The Anthropic API key must never appear in source code, Git commits, logs, or error messages.

---

18.3 Models

ITI uses the following Claude models:

Model	Use Case	Max Tokens (output)
claude-opus-4-5	Complex reasoning, long documents, multi-agent tasks	8,192
claude-sonnet-4-5	Balanced speed/quality for product features	8,192
claude-haiku-3-5	High-volume, latency-sensitive tasks	4,096

Default for products: claude-opus-4-5 unless performance requirements demand a faster model.

Note: Model names and capabilities change with Anthropic releases. Check operations/claude-parity-requirements.md for the current model selection guidance.

---

18.4 Basic API Call Structure

Python

import anthropic

client = anthropic.Anthropic()  # reads ANTHROPIC_API_KEY from env

message = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=2048,
    system="You are an expert travel planner.",
    messages=[
        {"role": "user", "content": "Plan a 7-day trip to Japan."}
    ]
)

print(message.content[0].text)

TypeScript (Node.js / Tauri)

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

const client = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY });

const message = await client.messages.create({
    model: 'claude-opus-4-5',
    max_tokens: 2048,
    system: 'You are an expert travel planner.',
    messages: [{ role: 'user', content: 'Plan a 7-day trip to Japan.' }],
});

console.log(message.content[0].text);

---

18.5 Tool Use (Function Calling)

Tool use allows Claude to request that the calling code execute a function and return the result. This is how multi-step reasoning is implemented without n8n.

import anthropic
import json

client = anthropic.Anthropic()

# Define tools
tools = [
    {
        "name": "search_flights",
        "description": "Search for available flights between two cities.",
        "input_schema": {
            "type": "object",
            "properties": {
                "origin": {"type": "string", "description": "Origin city or airport code"},
                "destination": {"type": "string", "description": "Destination city or airport code"},
                "date": {"type": "string", "description": "Travel date (YYYY-MM-DD)"},
            },
            "required": ["origin", "destination", "date"],
        },
    }
]

# Initial message
response = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    tools=tools,
    messages=[{"role": "user", "content": "Find me flights from Atlanta to Tokyo next week."}],
)

# Process tool calls
if response.stop_reason == "tool_use":
    tool_use_block = next(b for b in response.content if b.type == "tool_use")
    tool_result = execute_tool(tool_use_block.name, tool_use_block.input)

    # Continue conversation with tool result
    final_response = client.messages.create(
        model="claude-opus-4-5",
        max_tokens=1024,
        tools=tools,
        messages=[
            {"role": "user", "content": "Find me flights from Atlanta to Tokyo next week."},
            {"role": "assistant", "content": response.content},
            {"role": "user", "content": [
                {"type": "tool_result", "tool_use_id": tool_use_block.id, "content": json.dumps(tool_result)}
            ]},
        ],
    )

---

18.6 Streaming

For real-time UI updates (typing indicator effect), use streaming:

with client.messages.stream(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a poem about the ocean."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

---

18.7 Context Window Management

Claude has a context window (the maximum combined length of system prompt + conversation history + tool definitions + response). Exceeding the limit causes an error.

Model	Context Window
claude-opus-4-5	200,000 tokens
claude-sonnet-4-5	200,000 tokens
claude-haiku-3-5	200,000 tokens

Practical limits for product features:

• Keep system prompts under 2,000 tokens.
• Keep RAG-injected context under 8,000 tokens.
• Keep conversation history truncated to the last 20 messages or 10,000 tokens.

The ConversationManager in the ITI shared library handles automatic history truncation.

---

18.8 Error Handling

import anthropic

client = anthropic.Anthropic()

try:
    response = client.messages.create(...)
except anthropic.RateLimitError:
    # Wait and retry — use exponential backoff
    time.sleep(60)
except anthropic.APIConnectionError:
    # Network issue — retry or use fallback
    pass
except anthropic.APIStatusError as e:
    # API returned an error status
    print(f"API error: {e.status_code} {e.message}")
except anthropic.AuthenticationError:
    # Invalid API key — check configuration
    pass

---

18.9 Direct vs n8n-Mediated Calls

Consideration	Direct API	Via n8n
Centralized logging	No	Yes
RAG integration	Manual	Built-in (Dify retrieval node)
Error monitoring	Manual	Via Error Monitor workflow
Behavior changes without deployment	No	Yes (edit workflow)
Latency	Lower	Higher (webhook round-trip)
Suitable for	Fallback, scripts, prototypes	All product integrations

---

Previous: Chapter 17 — iOS & macOS with Swift | Next: Chapter 19 — Prompt Engineering