Latency Guide

Claude Slow Today? How to Diagnose Latency vs Outage

Latency incidents often look like outages. This guide helps you measure, verify, and decide when to reroute workloads.

Unknown

Live probe summary for Claude (updates every 60s on this page)

Latency

HTTP

Region

Checked

--:--:-- UTC

Fresh probeRefreshing…

Last updated: 2026-02-27

How to check if latency is platform-wide

Compare current latency signal on monitor to baseline.
Run short prompts instead of long jobs for quicker checks.
Check incident feeds for degraded performance notices.

What to do when Claude remains slow

Throttle non-critical usage.
Move priority tasks to faster backup assistant.
Retry later once status and latency normalize.

5-Minute Decision Matrix

Signal	Likely Meaning	Action Now
Claude unavailable for most users	Platform incident likely	Switch critical flows to backup and communicate ETA uncertainty.
Only specific account/session affected	Auth/session issue likely	Re-authenticate once and retry in private mode.
Claude slow but not fully down	Degraded performance	Throttle non-critical tasks and reroute priority tasks.

Incident Update Template (Copy/Paste)

[Incident update] Claude reliability issue

- Current impact: [blocked/degraded] workflows
- Scope: [team/service/region]
- Verification: live monitor + official status check completed
- Mitigation: switched critical tasks to backup assistant
- Next check-in: [UTC time in 5-10 minutes]