Is Claude Down? Complete Claude Status Guide + Troubleshooting (2026)

TLDR: Check Claude status instantly at apistatuscheck.com/api/anthropic or Anthropic's official page at status.anthropic.com. If Claude is down, use fallback providers (GPT-4o, Gemini). Common issues include API rate limits (429), overloaded errors (529), and Claude Code CLI authentication failures. This guide covers everything from quick status checks to building resilient multi-provider AI applications.

Claude stopped responding. Whether you're chatting on claude.ai, building with the Anthropic API, or running Claude Code in your terminal, an outage means your AI workflows grind to a halt. With Claude powering everything from enterprise coding assistants to research pipelines, knowing how to quickly diagnose and work around downtime is essential.

This guide covers every Claude product, common error codes, recent outage patterns, and production-ready fallback strategies to keep your AI features running when Anthropic has problems.

🔴 Is Claude Actually Down Right Now?

Claude spans multiple products that can fail independently. Check the right one:

Claude.ai (Consumer Chat)

1. API Status Check — Anthropic — Independent real-time monitoring across all Claude services
2. Anthropic Status Page — Official status from Anthropic
3. Downdetector — Claude — Community-reported outage spikes

Claude API (Developers)

1. API Status Check — Anthropic — Tracks API response times and availability
2. status.anthropic.com — Check "API" component specifically
3. Quick test from your terminal:

curl -s -o /dev/null -w "%{http_code}" https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-4-20250514","max_tokens":10,"messages":[{"role":"user","content":"hi"}]}'

• 200 = API is working
• 429 = Rate limited (not an outage — your quota)
• 500/502/503 = Server issues (likely an outage)
• 529 = Overloaded (Anthropic-specific — high demand)

Claude Code (CLI Tool)

1. API Status Check — Claude Code — Dedicated monitoring
2. Check Anthropic API status — Claude Code uses the same API backend
3. Common CLI-specific issues:
   • OAuth token expired → run claude login to re-authenticate
   • Local network/proxy blocking API calls
   • Outdated CLI version → npm update -g @anthropic-ai/claude-code

Understanding Anthropic's Architecture

These services share infrastructure but can fail independently:

• claude.ai — Consumer web app with its own capacity limits and session management
• Anthropic API — RESTful API for developers, separate rate limits and quotas per API key
• Claude Code — CLI tool that calls the Anthropic API (inherits API status)
• Claude for Enterprise — Dedicated capacity, separate from consumer
• Claude in AWS Bedrock — Runs on AWS infrastructure, independent from Anthropic's direct API
• Claude in Google Vertex AI — Runs on Google Cloud, independent from Anthropic's infrastructure

Key insight: If claude.ai is down but you need Claude urgently, the API often still works (and vice versa). AWS Bedrock and Google Vertex AI run on entirely separate infrastructure — they're your best bet during Anthropic-side outages.

Common Claude Errors and How to Fix Them

Error 529: Overloaded (Anthropic-Specific)

This is Claude's signature error — unique to Anthropic. It means their servers are handling too many requests.

What it means: Claude's inference capacity is maxed out. This is NOT a rate limit on your account — it's a global capacity issue.

How to fix:

1. Wait 30-60 seconds and retry — these usually resolve quickly
2. Implement retry with jitter:

import time
import random

def call_claude_with_retry(prompt, max_retries=5):
    for attempt in range(max_retries):
        try:
            return client.messages.create(
                model="claude-sonnet-4-20250514",
                max_tokens=4096,
                messages=[{"role": "user", "content": prompt}]
            )
        except anthropic.OverloadedError:
            if attempt == max_retries - 1:
                raise
            # Exponential backoff with jitter
            delay = (2 ** attempt) + random.uniform(0, 1)
            print(f"Claude overloaded, retrying in {delay:.1f}s...")
            time.sleep(delay)

3. Try a different model — Claude Haiku has separate capacity from Sonnet/Opus
4. Switch to AWS Bedrock or Google Vertex — separate infrastructure

Error 429: Rate Limited

What it means: You've exceeded YOUR account's rate limits, not a global outage.

Anthropic API rate limits (as of 2026):

• Free tier: 5 RPM, 25K input tokens/min, 5K output tokens/min
• Build tier ($5 deposit): 50 RPM, 500K input tokens/min, 50K output tokens/min
• Scale tier: 1,000 RPM, 4M input tokens/min, 400K output tokens/min
• Enterprise: Custom limits

How to fix:

1. Check your current usage in the Anthropic Console
2. Implement token-aware rate limiting:

class ClaudeRateLimiter {
  constructor(maxRPM = 50) {
    this.requests = [];
    this.maxRPM = maxRPM;
  }

  async waitForSlot() {
    const now = Date.now();
    this.requests = this.requests.filter(t => now - t < 60000);
    
    if (this.requests.length >= this.maxRPM) {
      const oldestRequest = this.requests[0];
      const waitTime = 60000 - (now - oldestRequest);
      console.log(`Rate limit approaching, waiting ${waitTime}ms`);
      await new Promise(r => setTimeout(r, waitTime));
    }
    
    this.requests.push(Date.now());
  }
}

3. Upgrade your tier — deposit $5 for Build tier (10x limits)
4. Batch requests when possible instead of many small calls

Error 500/502/503: Server Errors

What it means: Anthropic's servers are having issues. This IS an outage.

How to fix:

1. Check status.anthropic.com to confirm
2. Don't hammer the API with retries — use exponential backoff
3. Activate your fallback provider (see Alternatives section below)
4. These typically resolve in 15-60 minutes

"Claude is at capacity right now" (Web App)

What it means: The claude.ai consumer app is throttling new sessions.

How to fix:

1. Wait 5-10 minutes and refresh
2. Use the API instead — has separate capacity pools
3. Try during off-peak hours (late night US time)
4. Claude Pro subscribers get priority access

Error 401: Authentication Failed

What it means: Your API key is invalid, expired, or missing.

How to fix:

1. Check your API key:

echo $ANTHROPIC_API_KEY  # Should start with sk-ant-

2. Generate a new key at console.anthropic.com/settings/keys
3. For Claude Code: Run claude login to refresh OAuth
4. Check billing status — API access requires active billing

"Prompt is too long" / Context Length Exceeded

What it means: Your input exceeds the model's context window. NOT an outage.

Context limits (2026):

• Claude 3 Haiku: 200K tokens
• Claude 3.5 Sonnet: 200K tokens
• Claude Sonnet 4: 200K tokens
• Claude Opus 4: 200K tokens

How to fix:

1. Trim conversation history — keep only recent messages
2. Summarize long documents before sending
3. Use the system parameter efficiently (counts toward context)

Claude Code CLI: Specific Troubleshooting

Claude Code has become essential for developers. Here's how to diagnose CLI-specific issues:

"Authentication failed" or "Token expired"

# Re-authenticate
claude login

# Check current auth status
claude config list

# If using API key directly
export ANTHROPIC_API_KEY="sk-ant-your-key-here"

"Connection refused" or Timeout Errors

# Test API connectivity
curl -v https://api.anthropic.com/v1/messages 2>&1 | grep "Connected to"

# Check for proxy/VPN interference
echo $HTTP_PROXY $HTTPS_PROXY

# Update Claude Code
npm update -g @anthropic-ai/claude-code

High Latency / Slow Responses

• Check your model selection — Opus is slower than Sonnet, Haiku is fastest
• Reduce context — large codebases sent as context slow responses
• Check Anthropic API status — degraded performance shows up before full outages
• Try a different region if using Bedrock (us-east-1 vs eu-west-1)

Recent Claude Outage History (2026)

Understanding outage patterns helps predict future issues.

Outage Frequency

Anthropic has experienced notable incidents throughout early 2026. Based on our monitoring data, Claude API incidents have occurred roughly every 2-4 days, ranging from brief degradations (15-30 minutes) to longer outages (2-4 hours). Most incidents affect the API and claude.ai simultaneously, though the consumer app sometimes recovers faster.

Notable 2026 Incidents

• March 19, 2026 — API degradation affecting Claude Sonnet 4, elevated 529 errors for approximately 2 hours during US afternoon peak
• March 17-18, 2026 — Intermittent API availability issues, response latency spikes up to 30 seconds
• March 13, 2026 — Brief API outage lasting approximately 45 minutes, affecting both claude.ai and API
• March 11-12, 2026 — Extended degradation across multiple Claude models, 529 errors widespread
• March 6, 2026 — API instability following model updates, resolved within 2 hours
• March 2-4, 2026 — Multiple consecutive days of intermittent issues, particularly affecting Opus workloads
• February 2026 — Frequent incidents (Feb 16, 17, 20, 21, 23-27) suggesting infrastructure scaling challenges during rapid growth
• February 16, 2026 — Major outage coinciding with high demand after Claude 3.5 Sonnet updates

Common Outage Patterns

Peak risk times:

• US business hours (9 AM - 5 PM PST) — highest API traffic
• After major model releases (increased demand spikes)
• Monday mornings (accumulated batch jobs from weekends)
• Anthropic has experienced clusters of incidents (2-4 days in a row), suggesting that root causes sometimes take multiple attempts to fully resolve

Typical outage progression:

1. Response times increase (latency warning) — often noticeable 15-30 minutes before full outage
2. 529 Overloaded errors spike — first sign for API users
3. 500/503 errors begin appearing — service is degraded
4. Full outage (all requests failing) — claude.ai shows error pages
5. Gradual recovery (some models first, then all) — Haiku typically recovers first, then Sonnet, then Opus

Average resolution times:

• Minor degradation (elevated latency only): 15-45 minutes
• API outage (500/503 errors): 1-3 hours
• Major incident (complete unavailability): 2-6 hours
• Multi-day instability clusters: Can persist for 2-5 days with intermittent recovery

How to Track Outage History

• API Status Check — Anthropic incidents — Complete incident timeline
• status.anthropic.com — Official incident reports
• RSS feed: Subscribe to https://apistatuscheck.com/feed/anthropic for instant notifications

Alternatives When Claude Is Down

Don't let a Claude outage stop your work. Here are battle-tested alternatives:

For Chat / General Use

• ChatGPT — GPT-4o is the closest in capability, great for coding and analysis
• Google Gemini — Strong multimodal capabilities, generous free tier
• Perplexity — Best for research with real-time web citations
• Microsoft Copilot — Free GPT-4 access via Bing

For Developers (API Alternatives)

• OpenAI API — GPT-4o Turbo, 128K context, largest ecosystem
• Google Gemini API — 1M token context, generous free tier (15 RPM)
• Groq — Ultra-fast inference (800+ tokens/sec), open-source models
• AWS Bedrock — Claude — Same Claude models on AWS infrastructure (separate from Anthropic outages)
• Google Vertex AI — Claude — Same Claude models on Google Cloud infrastructure

For Claude Code Users

• GitHub Copilot — IDE-integrated, great for inline completions
• Cursor — AI-first code editor with multi-model support
• Windsurf — Free AI coding with strong autocomplete
• OpenAI Codex / ChatGPT — Paste code and ask questions directly

Production Fallback Pattern

import anthropic
import openai
import google.generativeai as genai

class ResilientAI:
    """Multi-provider AI client with automatic failover."""
    
    def __init__(self):
        self.claude = anthropic.Anthropic()
        self.openai = openai.OpenAI()
        genai.configure(api_key=os.environ["GEMINI_API_KEY"])
    
    async def generate(self, prompt: str, system: str = "") -> dict:
        # Primary: Claude (best for analysis and coding)
        try:
            response = self.claude.messages.create(
                model="claude-sonnet-4-20250514",
                max_tokens=4096,
                system=system,
                messages=[{"role": "user", "content": prompt}]
            )
            return {"text": response.content[0].text, "provider": "claude"}
        except (anthropic.APIStatusError, anthropic.APIConnectionError) as e:
            print(f"Claude failed: {e}")
        
        # Fallback 1: GPT-4o
        try:
            response = self.openai.chat.completions.create(
                model="gpt-4o",
                messages=[
                    {"role": "system", "content": system},
                    {"role": "user", "content": prompt}
                ]
            )
            return {"text": response.choices[0].message.content, "provider": "gpt-4o"}
        except Exception as e:
            print(f"OpenAI failed: {e}")
        
        # Fallback 2: Gemini
        try:
            model = genai.GenerativeModel("gemini-2.0-flash")
            response = model.generate_content(prompt)
            return {"text": response.text, "provider": "gemini"}
        except Exception as e:
            raise RuntimeError(f"All AI providers failed. Last error: {e}")

Claude via AWS Bedrock: Your Outage Insurance

AWS Bedrock runs Claude on Amazon's infrastructure — completely independent from Anthropic's servers. During Anthropic outages, Bedrock often continues working.

import boto3
import json

bedrock = boto3.client("bedrock-runtime", region_name="us-east-1")

def claude_via_bedrock(prompt: str) -> str:
    """Call Claude through AWS Bedrock — separate from Anthropic infrastructure."""
    response = bedrock.invoke_model(
        modelId="anthropic.claude-sonnet-4-20250514-v1:0",
        body=json.dumps({
            "anthropic_version": "bedrock-2023-05-31",
            "max_tokens": 4096,
            "messages": [{"role": "user", "content": prompt}]
        })
    )
    result = json.loads(response["body"].read())
    return result["content"][0]["text"]

When to use Bedrock:

• Anthropic API is returning 529/503 errors
• You need guaranteed SLA (Bedrock has 99.9% uptime commitment)
• Enterprise compliance requirements (data stays in your AWS account)
• You want to avoid Anthropic rate limits entirely

Monitoring Claude Proactively

Set Up Alerts

1. Bookmark apistatuscheck.com/api/anthropic for instant status checks
2. Subscribe to RSS — https://apistatuscheck.com/feed/anthropic
3. Follow @AnthropicAI on X/Twitter for official updates
4. Subscribe to status.anthropic.com email notifications

Build Your Own Health Check

// Monitor Claude API availability from your infrastructure
async function checkClaudeHealth(): Promise<{
  status: 'operational' | 'degraded' | 'down';
  latencyMs: number;
  error?: string;
}> {
  const start = Date.now();
  
  try {
    const response = await fetch('https://api.anthropic.com/v1/messages', {
      method: 'POST',
      headers: {
        'x-api-key': process.env.ANTHROPIC_API_KEY!,
        'anthropic-version': '2023-06-01',
        'content-type': 'application/json',
      },
      body: JSON.stringify({
        model: 'claude-haiku-3-20240307', // Cheapest model for health checks
        max_tokens: 5,
        messages: [{ role: 'user', content: 'hi' }],
      }),
      signal: AbortSignal.timeout(15000),
    });
    
    const latencyMs = Date.now() - start;
    
    if (response.ok) {
      return {
        status: latencyMs > 5000 ? 'degraded' : 'operational',
        latencyMs
      };
    }
    
    if (response.status === 529) {
      return { status: 'degraded', latencyMs, error: 'Overloaded' };
    }
    
    return { status: 'down', latencyMs, error: `HTTP ${response.status}` };
  } catch (error: any) {
    return {
      status: 'down',
      latencyMs: Date.now() - start,
      error: error.message
    };
  }
}

The "Claude Is Down" Quick Checklist

1. ✅ Check apistatuscheck.com/api/anthropic — confirm the outage
2. ✅ Check status.anthropic.com — official Anthropic status
3. ✅ Identify which Claude product — Web app? API? Claude Code? Bedrock?
4. ✅ Check error code — 429 = your rate limit, 529 = global overload, 503 = outage
5. ✅ Try a different model — Haiku often has separate capacity from Sonnet/Opus
6. ✅ Try AWS Bedrock — separate infrastructure, often unaffected
7. ✅ Activate fallback — GPT-4o or Gemini for urgent work
8. ✅ For Claude Code — try claude login if auth errors, or switch to API key auth

Claude Plans and Reliability Tiers

Understanding which Claude plan you're on helps troubleshoot access issues:

Consumer Plans (claude.ai)

• Free: Limited messages per day, throttled during peak hours, first to see "at capacity" errors
• Claude Pro ($20/month): Higher usage limits, priority access during peak demand, access to latest models including Opus
• Claude Team ($25/user/month): Team collaboration features, higher per-user limits, admin controls
• Claude Enterprise: Custom pricing, dedicated capacity, SSO, audit logs, guaranteed SLA

API Tiers

• Free tier: 5 RPM, good for testing — but easily mistaken for an outage when you hit limits
• Build tier ($5 deposit): 50 RPM — the minimum for any serious development
• Scale tier: 1,000 RPM — production workloads
• Enterprise: Custom limits and dedicated infrastructure

Pro tip for reliability: If you're building production features, don't rely solely on Anthropic's direct API. Set up AWS Bedrock or Google Vertex AI as a secondary path to the same Claude models on infrastructure with formal SLA commitments.

FAQ

Is Claude down right now?

Check apistatuscheck.com/api/anthropic for real-time status. We monitor Claude's API, web app, and Claude Code endpoints every 60 seconds from multiple locations. If you see green status but still experience issues, the problem is likely with your specific account, API key, or network — see the troubleshooting sections above.

What's the difference between 429 and 529 errors?

A 429 error means YOUR account hit its rate limit — you're sending too many requests. A 529 error is Anthropic-specific and means their servers are globally overloaded. For 429, slow down or upgrade your tier. For 529, wait and retry with exponential backoff, or switch to a fallback provider.

Can I use Claude through AWS Bedrock when Anthropic is down?

Yes! AWS Bedrock runs Claude on Amazon's infrastructure, completely independent from Anthropic's servers. During Anthropic API outages, Bedrock typically continues working normally. You need an AWS account with Bedrock access enabled. The same Claude models are available, though model version updates may lag by a few days.

Why does Claude Code keep disconnecting?

Claude Code CLI issues are usually caused by: expired OAuth tokens (fix with claude login), outdated CLI version (fix with npm update -g @anthropic-ai/claude-code), corporate proxy/VPN interference (check $HTTP_PROXY), or Anthropic API outages (check status page). If the API is operational but Claude Code fails, try using an API key directly via environment variable instead of OAuth.

How often does Claude go down?

Based on our 2026 monitoring data, Claude experiences some form of degradation roughly every 2-4 days, though most incidents are brief (15-45 minutes). Full outages lasting more than an hour occur approximately 2-3 times per month. Anthropic's uptime has improved over time, but any production application should still implement fallback providers.

Is Claude Pro worth it for better reliability?

Claude Pro ($20/month) gives you priority access to claude.ai during high-demand periods, which reduces "at capacity" errors. However, it does not protect against API outages. For API users, upgrading from the free tier to the Build tier ($5 deposit) significantly increases rate limits (from 5 to 50 RPM). For production reliability, AWS Bedrock or Google Vertex AI provide the best uptime guarantees.

Last updated: March 2026. Status information monitored in real-time at API Status Check.