June 17, 2026·11 min read

Claude Code Status: Is It Down? Check Guide (2026)

Claude CodeStatusTroubleshootingGuide

Claude Code status: how to check if it is down

The fastest way to check Claude Code status is the official Anthropic status page at status.anthropic.com, which shows live component health and a running log of incidents. If that page reports an active incident affecting the API or Claude Code, the problem is on Anthropic's side and there is nothing to fix on your machine. If the page is all green but Claude Code is still failing for you, the issue is almost certainly local: your network, your auth, your config, or a rate limit.

That single distinction, service-side outage versus local problem, is the whole game. Most of the time people spend reinstalling, clearing caches, and re-authenticating during an outage is wasted, because the tool was never broken. This guide shows you how to read the status signal correctly, how to tell the two cases apart in under a minute, and what to actually do in each case.

If you are new to the tool itself, our introduction to Claude Code covers what it is and how it works. This page is purely about checking status and recovering when something is wrong.

The 30-second status check

Run these three checks in order. The first one that gives you a clear answer ends the investigation.

#	Check	Where	What it tells you
1	Official status page	status.anthropic.com	Is Anthropic reporting an incident right now?
2	Your terminal error	The actual error text Claude Code printed	The error code points straight at the cause
3	A second network or account	Phone hotspot, different machine	Isolates whether the problem is your local setup

Start with the status page. It is the authoritative source: it is published by Anthropic, it breaks status out by component (the API, the Claude apps, and Claude Code), and it carries an incident history plus any scheduled maintenance. Third-party "is it down" sites can be useful as a second opinion, but they infer status from signals and user reports, so they lag and occasionally cry wolf. Treat the official page as truth and the aggregators as corroboration.

How to tell an outage from a local issue

This is where most people go wrong. Here is the decision tree.

Step 1: Check the official status page. If status.anthropic.com shows an active incident on the API or Claude Code, stop. It is an outage. Skip to the "what to do during an outage" section below. You cannot fix a server-side problem from your terminal.

Step 2: Read the actual error. If the status page is green, read the exact error Claude Code printed. The HTTP status code in the error is the single most useful diagnostic you have:

Error / code	Most likely meaning	Outage or local?
500, 502, 503, 529 (server errors)	Anthropic-side failure or overload	Service-side (check status page)
429 (too many requests)	You hit your own rate limit	Local to your account, not an outage
401 / 403 (auth)	Expired, missing, or invalid credentials	Local (your auth)
Connection timeout / DNS failure	Your network cannot reach the API	Local (your network)
Works in chat, fails in Claude Code	CLI, config, or MCP problem	Local (your setup)

A 5xx error is the one case that genuinely points at Anthropic. Everything else is on your side. A 429 is the most commonly misread of all: it is not an outage, it means your plan's usage window is exhausted. We cover that case in depth in the Claude Code rate limit fix guide, because the recovery is completely different from waiting out an outage.

Step 3: Test a second path. Still unsure? Switch your machine to a phone hotspot, or try Claude Code on a different machine or network. If it works on the second path, the problem is your local network or environment, not Anthropic. If it fails identically everywhere and the status page is green, you have likely found an incident that has not been posted yet, give it a few minutes and refresh the status page.

A quick tell: if the Claude chat apps (claude.ai or the desktop chat app) are working fine but only Claude Code is failing, that points away from a full outage and toward something specific to the CLI, your terminal, or a connected MCP server. Remember the chat app and Claude Code are separate products that happen to share the same model backend.

What the official status page actually shows

The Anthropic status page is more detailed than a single up/down light. It breaks health out by component, so you can see whether the issue is the core API, the consumer Claude apps, or Claude Code specifically. It also publishes:

Current incidents with a short description and a timeline of updates as the team investigates and resolves
Component status (operational, degraded performance, partial outage, major outage)
Incident history, so you can see whether this is a one-off or part of a run of recent issues
Scheduled maintenance, so a planned window does not get mistaken for a surprise outage

"Degraded performance" is worth calling out. It does not mean Claude Code is down. It usually means requests are slower, or a slice of them are failing and retrying. During degraded periods, Claude Code often still works but feels sluggish or throws the occasional 5xx that succeeds on retry. That is the status page doing its job: telling you the wobble is real and on Anthropic's side, so you stop debugging your own setup.

One honest caveat: the status page is updated by humans responding to an incident, so there is sometimes a short lag between "users start seeing errors" and "the page goes red." If your error is a clear 5xx and a second network reproduces it, trust your own evidence even if the page has not caught up yet.

What to do during a confirmed outage

If the status page confirms an incident, your options are limited but sensible:

Wait and watch the incident timeline. The page posts updates as the fix progresses. Subscribe to status updates if you want a notification rather than refreshing.
Do not reinstall or re-authenticate. Nothing on your machine is broken. Reinstalling during an outage just adds variables and wastes time.
Retry transient 5xx errors. During a partial or degraded incident, many requests still go through. A failed request often succeeds on the second or third try.
Switch models if only one is affected. Incidents sometimes hit a single model. If your default model is erroring, a different model may be unaffected for the moment. Anthropic's current lineup is Opus 4.8, Sonnet 4.6, and Haiku 4.5, and they do not always fail together.
Fall back to a non-AI workflow for anything urgent, then pick the task back up once the incident clears.

Anthropic has periodically acknowledged stretches of heavy demand straining its infrastructure, which can show up as clusters of short incidents rather than one long outage. When that happens, the pattern on the status page is your best guide to whether to wait it out or pause for the day.

When the status page is green but Claude Code still fails

This is the more common situation, and the good news is it is fixable. If status.anthropic.com is all green, work through the local causes in rough order of likelihood:

Rate limit (429). You have used up your plan's allowance for the current window. This is not an outage and not a bug. Wait for the window to roll forward, or move to a higher tier if you hit it constantly. The full playbook is in Claude Code rate limit fix, and the Claude Code pricing guide explains which plan gives you the headroom to stop hitting it.

Authentication (401 / 403). Your login or API key has expired or is missing. Re-running the login flow, or checking that the right API key is set in your environment, resolves most auth failures.

Network or firewall. A corporate proxy, VPN, or firewall can block the API endpoint while the rest of your internet works fine. Testing on a phone hotspot isolates this instantly.

A broken MCP server. If Claude Code itself starts but a connected tool fails, the culprit is often an MCP server, not Claude Code. Those failures look like outages but are entirely local. See Claude Code MCP not connecting for the fix.

Local config or install. A malformed config file or a half-finished update can break startup. Our Claude Code not working guide walks through the common local breakages, and install Claude Code covers a clean reinstall if you genuinely need one.

The throughline: when the status page is green, resist the urge to assume Anthropic is broken. Read the error, match it to the table above, and fix the specific local cause.

How Claudify fits in

Status checking is reactive. The thing that actually reduces how often you are blocked is how well Claude Code is configured, because a well-configured setup wastes fewer requests and fails in cleaner, more diagnosable ways.

Claudify is a pre-built configuration stack for Claude Code: skills, persistent memory, specialist agents, and hooks, installed in one command. It does not change Anthropic's uptime, and it will not get you through a server-side outage. What it does is make your day-to-day usage more efficient, so you burn less of your rate-limit budget on context reconstruction and spend more of it on real work. That means fewer self-inflicted 429s, which are the most common "is it down?" false alarm. If you want that headroom built in, see what's included.

Frequently asked questions

How do I check Claude Code status?

Go to the official Anthropic status page at status.anthropic.com. It shows live component health for the API, the Claude apps, and Claude Code, plus any active incidents and their resolution timeline. If it reports an incident on the API or Claude Code, the problem is on Anthropic's side. If it is all green but Claude Code is still failing for you, the issue is local, check your error code, network, auth, and rate limit instead.

Is Claude Code down right now or is it just me?

Check two things. First, the official status page: if it shows an active incident, it is down for everyone. Second, your error code: a 5xx (500/502/503/529) points to a server-side problem, while a 429 (rate limit), 401/403 (auth), or a connection timeout points to a local cause. If you are still unsure, try Claude Code on a phone hotspot or a different machine, if it works there, the problem is your local network or setup, not Anthropic.

Why is Claude Code not working when the status page is green?

A green status page means the most likely causes are local. The usual suspects, in order: you hit your rate limit (429), your authentication expired (401/403), a firewall or VPN is blocking the API, or a connected MCP server is failing. Read the exact error Claude Code printed and match it to the cause. Our Claude Code not working guide covers the local fixes step by step.

Does a 429 error mean Claude Code is down?

No. A 429 means "too many requests", you have used up your plan's allowance for the current usage window. It is specific to your account, not a service outage, and the status page will usually be green when you see it. The fix is to wait for the window to reset or move to a plan with more headroom, not to wait out an incident. See Claude Code rate limit fix for the details.

Should I use a third-party "is Claude down" site?

They are fine as a second opinion, but the official status page should be your primary source. Third-party aggregators infer status from signals and user reports, so they can lag the real incident or report a problem that is actually local to a cluster of users. Use the official page to confirm whether there is a real incident, and treat the aggregators as corroboration rather than the source of truth.

How long do Claude Code outages usually last?

It varies by incident, so there is no fixed number. The honest answer is to watch the incident timeline on the status page, which posts updates as the team works the fix. Some periods produce a cluster of short, overlapping incidents rather than one long outage. If you see that pattern, retrying transient 5xx errors often gets requests through between blips, and the timeline tells you whether to wait it out or pause for the day.

Summary

Checking Claude Code status comes down to one decision: outage or local issue.

Start at the official status page (status.anthropic.com). It shows component-level health, active incidents, history, and scheduled maintenance.
Read your error code. A 5xx is server-side. A 429 is your rate limit. A 401/403 is auth. A timeout is your network.
Test a second network if you are still unsure. If it works elsewhere, the problem is local.
During a confirmed outage, wait, retry transient errors, and do not reinstall, nothing on your machine is broken.
When the page is green but it still fails, work the local causes: rate limit, auth, network, MCP, or config.

Get that diagnosis right and you stop wasting time. The biggest avoidable cause of "is it down?" panic is a self-inflicted rate limit, which a leaner, better-configured setup makes far rarer.