Not just Microsoft.
While Microsoft 365 24h+ outage (Outlook, Teams, Exchange Online) still draining queues, Claude went dark at front door.
First incident — login:
- Opened 09:58 UTC Nov 13 — “investigating issue with users authenticating to http://Claude.ai and Claude Code”
- Status: http://claude.ai major outage, Claude Code, http://platform.claude.com Console partial outage, Claude Cowork degraded.
- Downdetector India spike: 2,895 reports at ∼3:45pm IST vs baseline 2 — 3,800+ globally per http://Vietnam.vn tracker. US Eastern 3am spike.
- Cause: not disclosed. Auth service — even logged-in sessions kicked out because renewal fails, macOS Keychain token, ~/.claude/.credentials.json Linux.
- Fix deployed 11:05 UTC, monitoring, marked resolved 11:58 UTC — ∼2 hours.
Claude API stayed operational during this window, so Bedrock / GCP Vertex users unaffected.
Second incident — wider:
- 13:09 UTC same day — degraded performance http://claude.ai, Claude API, Claude Code.
- 14:03 UTC list expands to connectors, Claude Security, Managed Agents.
- 14:36 UTC partial recovery http://claude.ai + API, but connectors, Claude in Meetings, Managed Agents still degraded.
- 15:33 UTC update: scheduled tasks + routines can run 2+ hours late, some skipped; adding new connector temporarily unavailable.
- Resolved 15:38 UTC — ∼2.5 hrs. Hits paid API layer developers build on.
Microsoft connector — separate but simultaneous:
Third incident opened 17:52 UTC Nov 12 — Microsoft 365 connector failing most requests. Anthropic: "This issue is caused by a Microsoft 365 service incident." Mail, calendar, SharePoint, Teams chat, Claude for Outlook, Claude in Outlook all fail. Still ongoing at 22:23 UTC Nov 12.
OpenAI same — confirmed ChatGPT Microsoft connectors also failing.
Anthropic says login failures NOT blamed on Microsoft — press inference. Only connector incident officially Microsoft-caused.
Impact:
- Developers: Claude Code is agentic coding CLI — blocked at auth prevents builds.
- Enterprise: Cowork office agent down, console can't rotate keys.
- India user base heavy — Moneycontrol reported surge.
As of 06:48 IST Nov 14, Anthropic status shows both Friday incidents resolved, but Microsoft connector incident still open, Microsoft says no ETA yet full recovery.
Workaround earlier: use API keys provisioned before outage, or via cloud providers bypassing http://claude.ai login.
#Claude #Anthropic #ClaudeCode #Outage #Authentication #Downdetector #Microsoft365 #AI #Console #Cowork