Is Claude down right now? Read the status page in 30 seconds
Open status.claude.com, Anthropic's own status page, and read the row for the part of Claude you use. Four steps cover it.

- Find your part: claude.ai for the chat site and apps, Claude Code for the terminal tool, Claude API (api.anthropic.com) for anything built on it.
- Read the label next to it. The five labels are explained in the next section.
- Open the newest incident and read its top update. The page posts times in UTC.
- Check the stage: investigating, identified, monitoring or resolved.
As of 15 November 2026 the page lists ten components. Beyond those three it has Claude Console, Claude Cowork, Claude for Government and four cloud-platform rows, among them Claude Platform on AWS and Claude in Microsoft Foundry, so if you reach Claude through a cloud provider, read that row. This page is evergreen, so it never states what the status is today. Only the status page can. It also lets you stop checking by hand: as of 15 November 2026 it offers email, text message, Slack, Microsoft Teams, webhook, Atom and RSS updates. Our is Claude down checker shows the status page next to user reports if you want both in one place.
A green page isn't proof that nothing is wrong. Anthropic's help center says: "Capacity constraints are not outages. The system is functioning normally but managing high demand across all users." It adds that capacity issues "will not appear on our status page". So a busy-model message on a green page is real. It passes.
What do the labels on the Claude status page mean?
Each part of Claude carries one of five labels, and the label tells you whether to keep troubleshooting. The status page lists the five labels in its legend, read 15 November 2026. Anthropic doesn't print a definition beside them, so the middle column below is how we read them.
| Label on status.claude.com | How we read it | What to do |
|---|---|---|
| Operational | Anthropic has posted no fault for that part | Look at your own setup, steps below |
| Degraded Performance | It works, but slowly or with some errors | Retry, shorten the prompt, or pick another model |
| Partial Outage | Part of the service, or some users, fail | Retry in a few minutes, then switch app or vendor |
| Major Outage | The part is largely down | Stop troubleshooting and use another tool |
| Maintenance | Planned work | Wait for the window to end |
Incidents move through stages too. Reading the incident updates in the status page's feed, the stages run investigating, identified, monitoring and resolved. Monitoring means a fix is in and Anthropic is watching it, so errors can still return for a while. In the feed, 29 of the 50 incidents reached monitoring, and the median of those ran another 24 minutes before the resolved notice. None of the 29 dropped back to investigating or identified. The 90-day bars under each component use their own wording: major outage, partial outage, no downtime recorded on this day, and no data exists for this day.
Which parts of Claude are down most often?
Claude.ai is listed most, and the chat site, the API and Claude Code usually go down together. On 15 November 2026 we read the 50 incidents in the status page's incident feed, which run from 14 August to 14 November 2026, and counted the components each one listed.
| Component, 14 Aug to 14 Nov 2026 | Incidents that listed it (of 50) |
|---|---|
| claude.ai | 40 |
| Claude Cowork | 35 |
| Claude Code | 32 |
| Claude API (api.anthropic.com) | 28 |
| Claude Console (platform.claude.com) | 8 |
The overlap matters more than the counts. 26 of the 50 listed claude.ai, the API and Claude Code at the same time, and 6 listed claude.ai alone. Only 1 listed Claude Code without claude.ai. So if chat is broken, your terminal session is probably struggling too. If only chat is broken, an app that calls the API may be fine.
We measured each incident from its posted start to its resolved notice. The median came to 62 minutes. One more incident was still open when we read the feed, so it sits outside the chart.
How long do Claude outages last? Dated examples
Most last under three hours, and a few run for days, so the first thing to read in any incident is the title, because it names the models or parts that are affected. The table picks six incidents from the same feed, each with its posted title and length. All six are on the status page's history.
| Start (UTC) | Posted title | Posted length |
|---|---|---|
| 24 Oct 2026 | Elevated errors for Claude Opus 5.5 | 6 minutes |
| 27 Oct 2026 | Elevated errors for Claude Sonnet 5.5 (two incidents) | 41 and 42 minutes |
| 13 Nov 2026 | Elevated errors on claude.ai, Claude Code, and platform.claude.com | 120 minutes |
| 13 Nov 2026 | Elevated errors on claude.ai, the Claude API, and Claude Code | 149 minutes |
| 12 Nov 2026 | Elevated errors affecting the Microsoft 365 connector | 1,677 minutes, about 28 hours |
| 7 Oct 2026 | Elevated errors on platform.claude.com | 8,915 minutes, about 6 days |
Model incidents ran short in this sample, with 6, 41 and 42 minutes, while the connector and Console incidents ran for more than a day each. The feed also rates each incident. 27 of the 50 were rated major and 2 critical, 18 were minor and 3 carried no rating, so most outages on this page were not small ones. Even so, 42 of the 49 resolved incidents ended within three hours.
If the title names one model, such as Opus 5.5, switch to another Claude model and keep going. If it names claude.ai and Claude Code together, expect a longer wait. The two examples above ran 120 and 149 minutes.
Is Claude down again? How often incidents hit, and when
Claude had an incident about every other day, and a new one started a median 17 hours after the last. We took the start time of all 50 incidents in the incident feed, 14 August to 14 November 2026. They began on 34 of the 93 days. The longest quiet stretch between two starts was about 14 days.
| Incident start, UTC | Incidents (of 50) |
|---|---|
| 00:00 to 05:59 | 8 |
| 06:00 to 11:59 | 8 |
| 12:00 to 17:59 | 21 |
| 18:00 to 23:59 | 13 |
So "is Claude down this morning" has a pattern behind it. The 12:00 to 17:59 UTC window is the US morning and the European afternoon, and it held 21 of the 50 starts. Wednesday was the busiest weekday with 11, and weekends held 8. Start times are the ones Anthropic posted, which can trail the first user errors, so treat the hours as a tendency, not a forecast.
Is Claude down for everyone, or is it just you?
Read the message first, because it says whose problem it is. A message with a reset time is your usage limit, not an outage. A capacity or overloaded message is Anthropic's side but never reaches the status page. Claude not working has a table that matches each wording to its fix.
Then test from outside your usual setup, which takes about a minute and rules out most faults in your own browser or network. Load claude.ai on your phone over mobile data with wifi off, or open the desktop app. If Claude works there, the fault is in your browser or network, whatever a crowd-sourced tracker says. Trackers, and threads on Reddit or X, count complaints, and complaints spike on any bad minute.
Why is Claude not working?
Three different things hide behind "not working", and the status page tells you about only one. An incident listed for your part is a Claude outage. A capacity or busy-model message is on Anthropic's side too, but it never reaches the status page and is not counted as an outage. A usage-limit message, a login failure or a page that won't load is yours to fix.
Ask two questions in order. Does status.claude.com list an incident for the part you use? If yes, wait it out or switch, as the table of labels above says. If no, does Claude work on your phone over mobile data? If it does, the fault is in your browser, VPN or extensions. Claude not working matches each error message to its fix, so this page stops at deciding whose fault it is.
What do you do when the status page is green?
Work through these in order, and stop at the first one that helps. The login steps are Anthropic's, from its error message guide.
- Copy your unsent message out of the box before you touch anything.
- Retry once after a minute. A capacity message usually passes within minutes.
- Pick a different Claude model in the model picker. Capacity is tracked per model.
- Try the same chat in another browser, the desktop app, or your phone.
- "Ensure you're not using a VPN when accessing Claude."
- "Disable any browser extensions that you currently have active."
- "Clear your browser's cache and cookies."
- Wait ten minutes and check the status page again, in case an incident has been posted since.
A repeated 500 error with no incident posted has its own page, Claude internal server error. A busy-model message has Claude overloaded.
Is Claude Code down, or is it your setup?
Check the Claude Code row on the status page, then run /doctor inside Claude Code. Anthropic's troubleshooting page says /doctor checks your install, settings, extensions and context usage and proposes fixes. If claude won't start at all, run claude doctor from your shell instead.
The error text then tells you which side broke. API Error: Repeated 529 Overloaded errors means the model is at capacity. Anthropic says a 529 isn't your usage limit and doesn't count against your quota, and /model lets you switch, since capacity is tracked per model. API Error: 500 Internal server error is a fault inside the API, not your prompt or account.
Unable to connect to API points at your network. Run curl -I https://api.anthropic.com from the same shell. On Windows PowerShell, run curl.exe -I https://api.anthropic.com instead. If you sit behind a corporate proxy, set HTTPS_PROXY before you launch Claude Code. A request that times out waits 10 minutes by default, and API_TIMEOUT_MS raises that.
Don't keep retrying by hand. Claude Code "retries transient failures up to 10 times with exponential backoff before showing you an error", so the error on screen has already been tried several times. To pick a stopped session back up later, run claude --resume. The full list is in the Claude Code error reference.
What can you use while Claude is down?
Another Claude model if the incident names one model, and another vendor if it names claude.ai as a whole. Copy the thread first. For the other vendors and their status pages, see what to use when ChatGPT is down, and AI chatbot outages compares how often each vendor posts incidents. Is Gemini down covers Google's three status pages.
Whizi, which publishes this page, carries Claude, GPT and Gemini in one app on Pro, and its first week is $1.99. GPT and Gemini are other companies' models, so a Claude incident doesn't stop work there, and the thread goes with you when you switch (how switching works). The limit: a Claude model in Whizi still runs on Anthropic's servers, so it isn't a way around an Anthropic outage.
Frequently asked questions
Is Claude down right now?
Check status.claude.com, because this page can't tell you today's state. It lists claude.ai, the Claude API, Claude Code and the Console, each with one of five labels. In its last 50 incidents, from 14 August to 14 November 2026, the median took 62 minutes to resolve. Claude not working covers what to do when the page is green.
Why does Claude say it is at capacity when the status page is green?
Because Anthropic doesn't post capacity issues there. Its help center says "Capacity issues will not appear on our status page because they represent normal load management." Retry after a few minutes or pick another Claude model. Claude overloaded lists the fixes in order.
Is Claude Code down when claude.ai is down?
Usually, but not always. In the last 50 incidents, 40 listed claude.ai and 32 listed Claude Code, and only 1 listed Claude Code without claude.ai. Check the Claude Code row on status.claude.com, then run /doctor to rule out your own setup.
How do I know if Claude is down or it is just me?
Open status.claude.com, then load claude.ai on your phone over mobile data. If the page is green and it works on your phone, your browser, VPN or extensions are the cause. Anthropic's fix is to turn off the VPN and extensions and clear the cache and cookies.
Is Claude down again?
Incidents are frequent: status.claude.com posted 50 from 14 August to 14 November 2026, starting on 34 of 93 days, with a median 17 hours between starts. 21 of the 50 began between 12:00 and 17:59 UTC. Check the status page for today, then follow the order of fixes above if it is green.