Anthropic Claude API Not Working? Fix Auth, Rate Limits & SDK Errors

401 Unauthorized / invalid API key

The most common cause is a missing, expired, or incorrectly formatted API key. Use the official SDK with an environment variable:

import anthropic
import os

client = anthropic.Anthropic(
    api_key=os.environ.get("ANTHROPIC_API_KEY")  # sk-ant-api03-...
)

message = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Hello"}
    ]
)
print(message.content[0].text)

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

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

const message = await client.messages.create({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Hello' }],
});

• Key format: all Anthropic API keys start with sk-ant-api03- — verify this prefix exactly.
• Get or rotate keys: go to console.anthropic.com/settings/keys — deleted or expired keys show as inactive.
• Claude.ai ≠ API: a Claude.ai subscription does NOT include API access — you need a separate API account at console.anthropic.com with billing enabled.
• Workspace keys: if using workspace-scoped keys, ensure the workspace has API credits and the key has not been restricted.

429 Rate limit — RPM, TPM, concurrent

Anthropic enforces three separate rate limit types simultaneously: requests per minute (RPM), tokens per minute (TPM, input + output combined), and concurrent requests. The Anthropic SDK has built-in retry logic you can configure:

# The Anthropic SDK has built-in retry for 429 and 529
client = anthropic.Anthropic(
    api_key=os.environ["ANTHROPIC_API_KEY"],
    max_retries=5,  # default is 2
)

# Or manual retry with backoff
import time
import random

def create_with_retry(client, **kwargs):
    for attempt in range(5):
        try:
            return client.messages.create(**kwargs)
        except anthropic.RateLimitError as e:
            if attempt == 4:
                raise
            wait = (2 ** attempt) + random.random()
            print(f"Rate limited. Retry {attempt + 1}/5 in {wait:.1f}s")
            time.sleep(wait)

• Check rate limit headers: inspect x-ratelimit-limit-requests, x-ratelimit-remaining-requests, and x-ratelimit-reset-requests to see which limit you hit.
• Tier progression: Anthropic tiers by historical spend — Tier 1 unlocks after a $5 payment; Tier 4 after $40k spent. Check your tier at console.anthropic.com/settings/limits.
• Concurrent limit: easy to hit in parallel code — use asyncio.Semaphore or a queue to cap simultaneous in-flight requests.

529 Overloaded — Anthropic server issue

Status 529 means Anthropic's servers are overloaded — it is a temporary server-side issue, not your code. Always retry 529 with exponential backoff:

# 529 should always be retried — it's a temporary server issue
from anthropic import APIStatusError

def robust_create(client, **kwargs):
    for attempt in range(6):
        try:
            return client.messages.create(**kwargs)
        except APIStatusError as e:
            if e.status_code in (429, 529) and attempt < 5:
                wait = min(60, (2 ** attempt) + random.random())
                time.sleep(wait)
            else:
                raise

Wrong model name

Claude model IDs are exact strings — any typo or wrong format returns a model-not-found error. Current valid model IDs (as of June 2026):

• claude-opus-4-8 — most capable
• claude-sonnet-4-6 — balanced performance
• claude-3-5-sonnet-20241022 — previous gen, stable
• claude-3-5-haiku-20241022 — fastest, cheapest
• claude-3-opus-20240229 — previous Opus

• Common wrong names that fail: claude-3.5-sonnet (dot instead of dash), claude-opus-4 (missing version suffix), claude-3-5-sonnet (missing date suffix for older models).
• Always check: console.anthropic.com/docs/models for the current canonical model IDs.

Streaming with the SDK

Use the stream() context manager and iterate — do not call message.content on a stream object before it completes:

# Python streaming
with client.messages.stream(
    model="claude-3-5-sonnet-20241022",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a poem"}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

# Get the final message after streaming
final_message = stream.get_final_message()
print(f"\nUsage: {final_message.usage}")

// TypeScript streaming
const stream = await client.messages.stream({
  model: 'claude-3-5-sonnet-20241022',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Write a poem' }],
});

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

• Common mistake: calling message.content on a stream object before it completes — always use the stream iterator or await stream.finalMessage().
• Usage stats: token counts are available via stream.get_final_message().usage (Python) or await stream.finalMessage() (TypeScript) after the stream completes.