Anthropic's status page showed 99.54% uptime for claude.ai over the 90 days to November 13, 2026, with the API at 99.77% and Claude Code at 99.69%. The worst month of the year was March, at 98.21% for claude.ai, which works out to about 13 hours of downtime. On November 13, Claude had two major incidents, from 09:58 to 11:58 UTC and from 13:09 to 15:38 UTC. Anthropic has not published a cause for either.
Uptime by Product
| Product | Past 30 days | Past 60 days | Past 90 days |
|---|---|---|---|
| claude.ai | 99.21% | 99.51% | 99.54% |
| Claude API | 99.79% | 99.82% | 99.77% |
| Claude Code | 99.73% | 99.78% | 99.69% |
| Claude Console | 99.95% | 99.78% | 99.82% |
| Claude Cowork | 99.87% | 99.85% | 99.74% |
| Claude on cloud platforms | 100% | 100% | 100% |
Uptime by Month, 2026
| Month | claude.ai | Claude API | Claude Code |
|---|---|---|---|
| January | 99.51% | 99.54% | 99.51% |
| February | 98.93% | 99.24% | 99.76% |
| March | 98.21% | 98.32% | 98.57% |
| April | 98.87% | 99.48% | 99.25% |
| May | 99.50% | 99.55% | 99.50% |
| June | 99.54% | 99.58% | 99.46% |
| July | 99.01% | 99.14% | 99.09% |
| August | 99.60% | 99.73% | 99.60% |
| September | 99.70% | 99.73% | 99.70% |
| October | 99.85% | 99.86% | 99.85% |
| November, to the 13th | 98.43% | 99.77% | 99.66% |
Reliability improved from August to October, then November's incidents pulled claude.ai back below 99%.
November 12 and 13, 2026
| Time (UTC) | Rating | What happened |
|---|---|---|
| Nov 12, from 17:52 | Major | Errors in the Microsoft 365 connector on claude.ai and Cowork. Anthropic said the cause was a Microsoft 365 service incident. Still being monitored on November 13 |
| Nov 13, 09:58 to 11:58 | Major | Users could not authenticate to claude.ai and Claude Code. A fix was deployed at 11:05 |
| Nov 13, 13:09 to 15:38 | Major | Elevated errors on claude.ai, the API and Claude Code, then connectors, Managed Agents, Claude Meetings and the mobile apps. Some scheduled tasks ran more than two hours late or were skipped |
The second incident ran from 8:09 to 10:38 a.m. Eastern, which matches the rise in US searches for "claude outage" that morning. For the Microsoft side, see the Microsoft 365 outage timeline.
Main Critical Incidents in 2026
"Critical" is Anthropic's own highest rating. Anthropic posted 18 critical incidents in 2026. This table covers 15 of them, and the model suspension described below is another. Two more, one in March and one in April, are not listed. Times are UTC.
| Date | Time | What was affected |
|---|---|---|
| Jan 14 | 08:34 to 14:17 | Opus 4.5 and Sonnet 4.5 errors on claude.ai, the API and Claude Code |
| Feb 3 | 15:46 to 15:56, again 17:52 to 17:56 | API errors across all models |
| Mar 22 | 18:31 to 18:51 | claude.ai and login down |
| Apr 7 | 14:32 to 15:59 | claude.ai login, voice mode and chats |
| Apr 13 | 15:40 to 16:35 | claude.ai down |
| Apr 15 | 14:53 to 17:42 | claude.ai and the platform down, Claude Code login failing |
| Apr 30 | 01:20 to 01:51 | claude.ai and API unavailable |
| Jun 23 | 14:19 to 16:44 | Errors across multiple models |
| Jul 6 | 19:18 to 20:05 | claude.ai, including Claude Code login |
| Jul 21 | 15:35 to 18:18 | Errors on several models, plus a wider service disruption |
| Jul 29 | 19:49 to 22:36 | Errors across all models |
| Aug 16 | 21:58 to 22:34 | Authentication, then claude.ai, console, API, Claude Code and Cowork |
| Oct 23 | 23:37 to 00:11 | Fable 5.1, Sonnet 5.5 and Haiku 5.5 |
One critical entry was not a technical outage. From June 13 to July 1, Anthropic suspended access to Claude Mythos 5 and Claude Fable 5 for all customers after a US government directive, and restored Fable 5 on July 1 when the controls were lifted. Other models were not affected.
Incident Posts Per Month
| Month | All posts | Critical | Major |
|---|---|---|---|
| January | 26 | 1 | 5 |
| February | 58 | 2 | 13 |
| March | 51 | 2 | 19 |
| April | 49 | 5 | 11 |
| May | 41 | 0 | 17 |
| June | 47 | 1 | 8 |
| July | 55 | 5 | 12 |
| August | 25 | 1 | 7 |
| September | 13 | 0 | 8 |
| October | 17 | 1 | 10 |
| November, to the 13th | 5 | 0 | 3 |
| Total | 387 | 18 | 113 |
Posts include small, model-specific notices. A post is not a measure of downtime, so use the uptime tables above for that.
Causes Anthropic Has Published
| Date | Cause, paraphrased from Anthropic's postmortem |
|---|---|
| Jan 12 to 13 | A configuration mismatch between inference servers made load balancing uneven and overloaded part of the fleet |
| Jan 14 | A service deployment temporarily reduced serving capacity |
| Jan 21 | Not enough compute capacity allocated to meet demand |
| Feb 3 | An internal configuration update, which was reverted |
| Mar 11 | The main application database slowed sharply after routine maintenance. The API was unaffected |
| Mar 26 to 27 | Network degradation between components of the serving stack |
| Aug 28 | An issue with an upstream cloud provider |
| Sep 10 to 14 | A Windows update left Cowork on Windows unable to run local commands |
Anthropic posted eight postmortems on its status page in 2026, all between January and March. We found none for April through November.
How to Check If Claude Is Down
- Official status: status.claude.com lists each product separately.
- Check which product: claude.ai, the API and Claude Code often fail independently. On March 11 the app was down and the API was not.
- Check login first: many 2026 incidents were authentication failures while the models kept working.
For a comparison with other assistants, see the AI assistant outage tracker.
Methodology
Uptime figures, incident times and ratings come from status.claude.com and its incident history, read on November 13, 2026 at about 17:19 UTC. Monthly uptime is rounded to two decimals. The downtime estimate for March is our arithmetic from the published percentage. Incident counts are our tally of all 2026 posts. The model suspension is described in Anthropic's June 12 notice. Presenc AI did not measure uptime independently.