anthropic-sdk-patterns
Apply production-ready Anthropic SDK patterns for TypeScript and Python. Use when implementing Anthropic integrations, refactoring SDK usage, or establishing team coding standards for Anthropic. Trigger with phrases like "anthropic SDK patterns", "anthropic best practices", "anthropic code patterns", "idiomatic anthropic".
claude-code
Allowed Tools
ReadWriteEdit
Provided by Plugin
anthropic-pack
Claude Code skill pack for Anthropic (30 skills)
Installation
This skill is included in the anthropic-pack plugin:
/plugin install anthropic-pack@claude-code-plugins-plus
Click to copy
Instructions
Anthropic SDK Patterns
Overview
Production-ready patterns for Anthropic SDK usage in TypeScript and Python.
Prerequisites
- Completed
anthropic-install-authsetup - Familiarity with async/await patterns
- Understanding of error handling best practices
Instructions
Step 1: Implement Singleton Pattern (Recommended)
// src/anthropic/client.ts
import { AnthropicClient } from '@anthropic/sdk';
let instance: AnthropicClient | null = null;
export function getAnthropicClient(): AnthropicClient {
if (!instance) {
instance = new AnthropicClient({
apiKey: process.env.ANTHROPIC_API_KEY!,
// Additional options
});
}
return instance;
}
Step 2: Add Error Handling Wrapper
import { AnthropicError } from '@anthropic/sdk';
async function safeAnthropicCall<T>(
operation: () => Promise<T>
): Promise<{ data: T | null; error: Error | null }> {
try {
const data = await operation();
return { data, error: null };
} catch (err) {
if (err instanceof AnthropicError) {
console.error({
code: err.code,
message: err.message,
});
}
return { data: null, error: err as Error };
}
}
Step 3: Implement Retry Logic
async function withRetry<T>(
operation: () => Promise<T>,
maxRetries = 3,
backoffMs = 1000
): Promise<T> {
for (let attempt = 1; attempt <= maxRetries; attempt++) {
try {
return await operation();
} catch (err) {
if (attempt === maxRetries) throw err;
const delay = backoffMs * Math.pow(2, attempt - 1);
await new Promise(r => setTimeout(r, delay));
}
}
throw new Error('Unreachable');
}
Output
- Type-safe client singleton
- Robust error handling with structured logging
- Automatic retry with exponential backoff
- Runtime validation for API responses
Error Handling
| Pattern | Use Case | Benefit |
|---|---|---|
| Safe wrapper | All API calls | Prevents uncaught exceptions |
| Retry logic | Transient failures | Improves reliability |
| Type guards | Response validation | Catches API changes |
| Logging | All operations | Debugging and monitoring |
Examples
Factory Pattern (Multi-tenant)
const clients = new Map<string, AnthropicClient>();
export function getClientForTenant(tenantId: string): AnthropicClient {
if (!clients.has(tenantId)) {
const apiKey = getTenantApiKey(tenantId);
clients.set(tenantId, new AnthropicClient({ apiKey }));
}
return clients.get(tenantId)!;
}
Python Context Manager
from contextlib import asynccontextmanager
from anthropic import AnthropicClient
@asynccontextmanager
async def get_anthropic_client():
client = AnthropicClient()
try:
yield client
finally:
await client.close()
Zod Validation
import { z } from 'zod';
const anthropicResponseSchema = z.object({
id: z.string(),
status: z.enum(['active', 'inactive']),
createdAt: z.string().datetime(),
});
Resources
Next Steps
Apply patterns in anthropic-core-workflow-a for real-world usage.