Is Claude down right now?
The live status from status.claude.com is below. During an outage, Claude API requests fail with HTTP 529 "overloaded_error" or HTTP 500 "api_error". Retry with backoff, switch model in your agent, or reach the same Claude model on Google Vertex AI or Amazon Bedrock.
Live Claude status
Checked
All systems operational
- Claude API (api.anthropic.com)Operational
- Claude CodeOperational
- claude.aiOperational
- Claude Console (platform.claude.com)Operational
- Last incident: Elevated errors on platform.claude.comresolved Oct 13
Source: status.claude.com. Read on our server about every 2 minutes.
Which error is Anthropic, and which is you
| What you see | What it means |
|---|---|
HTTP 529, overloaded_error | Anthropic's API is over capacity. See the 529 fix |
HTTP 500, api_error | An internal error at Anthropic. Retry; if it repeats for minutes, expect an incident on the status page |
HTTP 429, rate_limit_error | Your organization's own limit. See the rate limit fix |
HTTP 401, authentication_error | Your API key |
Claude Code shows API Error: 529 | The same 529. Claude Code retries on its own before it shows the error |
Five minutes into a Claude outage
- Retry with exponential backoff rather than in a tight loop. A retry storm makes overload worse for everyone, including you.
- Read the incident title on the status page. If it names one model, switch to another Claude model.
- If the whole Claude API is failing, switch your agent to another vendor's model with the steps below.
- Or keep Claude and change the provider: Google Vertex AI and Amazon Bedrock run Claude on their own infrastructure, so an outage of api.anthropic.com does not automatically reach them.
How to switch model in your agent
| Agent | Steps |
|---|---|
| Claude Code | Type /model and pick Opus, Sonnet or Haiku. For Bedrock or Vertex AI, see Is Claude Code down? |
| Cline, Roo Code, Kilo Code | Open the extension's settings, change API Provider or Model, and save. The next request uses it |
| OpenCode | Type /models and pick a model from another provider, or start with opencode -m provider/model |
| Aider | Type /model followed by another model name; the chat history stays |
| Your own code | Catch 5xx and 529, retry with exponential backoff, then send the request to a fallback model |
| Cursor | Pick a non-Claude model in the chat panel's model menu |
What Standard Compute does when Anthropic fails
- Every request goes through OpenRouter. Depending on the key, a Claude request goes to Anthropic's API or to Google Vertex AI first; if that provider fails, OpenRouter can send it to another provider of the same Claude model, such as Amazon Bedrock.
- Each failed call is retried twice on a 429, a 5xx or a network error, after 0.5 and then 1 second, or after the provider's Retry-After up to 8 seconds. If every attempt fails, your agent gets HTTP 503 with "The model is temporarily unavailable. Please retry shortly."
- A specific model ID always stays that model; failover only changes the provider. Only the auto-router, the model
standardcomputewith no model selection, can switch to a different model. - A key restricted to Europe or the US fails over only between the endpoints approved for that region.
The catch
Failover needs a provider that still works. If Claude fails on Anthropic, Vertex AI and Bedrock at the same time, Standard Compute cannot serve Claude either; switch to another model on the same key, such as openai/gpt-6-astra or z-ai/glm-5.3, or wait.
Where Standard Compute fits
Standard Compute is an OpenAI- and Anthropic-compatible LLM gateway for coding agents at https://api.stdcmpt.com/v1, with monthly plans from $19 (USD) and a fixed compute budget. One key serves Claude, GPT, Gemini, DeepSeek and GLM models, so moving off a struggling model during an outage takes one changed model ID.
FAQ
Is Claude down or is it just me?
Check status.claude.com (status.anthropic.com redirects there): the Claude API row covers api.anthropic.com and the Claude Code row covers Claude Code. If both are operational and requests still fail, read the status code: 401 means your key, 429 your own rate limit, and 529 Anthropic's capacity.
What does 529 overloaded_error mean on the Claude API?
Anthropic's API is temporarily over capacity. It is a capacity problem at Anthropic, and your billing and rate limits are fine. Retry with exponential backoff, and keep a fallback model or provider for longer spikes.
Does Claude on Amazon Bedrock or Google Vertex AI keep working when Anthropic is down?
They can, because they run Claude on their own infrastructure. Each has its own capacity and status page, and an incident can still affect more than one provider, so check theirs too.
Can a gateway keep Claude working during an Anthropic outage?
Only while another provider of the same Claude model works. On Standard Compute a Claude request can fail over between Anthropic, Google Vertex AI and other providers of that model, with two retries, and returns HTTP 503 if every attempt fails. A specific model ID always stays that model.
Does a gateway make Claude slower?
Yes, a little. Production logs for October 13, 2026 put the gateway's own work before a request leaves (key, budget and policy checks) at a median of about 0.6 seconds, and about 1.1 seconds when a request waits for the auto-router's model choice. Send a specific model ID to skip the router.