Elevated errors on Opus 4.5
Claude API (api.anthropic.com)
Check whether slow Claude replies, delayed Claude Code runs, usage-limit symptoms, timeouts, or sluggish API responses match a recent incident or a local setup issue.
Fast answer
If Claude is slow but not fully down, separate the symptom into service latency, account limits, local Claude Code context, or API errors before changing prompts or retry logic.
Slow across Claude.ai, API, and Claude Code, especially after a recent incident.
Requests still work, but responses feel throttled, delayed, or model access changes.
Slow only in one repository, terminal, long session, MCP setup, or large context.
Request timed out, stream idle timeout, 429, 529 overloaded, 500, or 503 errors need log-based triage.
A green status page does not always mean every Claude workflow is fast. Claude can be officially operational while a specific model, account, browser session, region, prompt, API route, or Claude Code run is still slow.
Use this checklist when Claude Code feels slow, degraded, lazy, worse than usual, or takes forever. The cause may be a service incident, but it may also be usage limits, long context, a large codebase, model routing, or too many tool calls.
If Claude Code is slow only in one terminal, OS, repository, or long-running session, treat it as a local workflow diagnosis before treating it as a full Claude service incident.
Latency can appear on Claude.ai, Claude API, Claude Code, Console, or only in a local browser, network, account, or prompt workflow.
| Component | Status | What it affects |
|---|---|---|
| claude.ai | operational | Web app, chat UI, and login experience. |
| Claude Console (platform.claude.com) | operational | Developer console and platform settings. |
| Claude API (api.anthropic.com) | operational | API requests to api.anthropic.com. |
| Claude Code | operational | Coding agent, CLI, and IDE-related workflows. |
| Claude Cowork | operational | Team collaboration features. |
| Claude for Government | operational | Government environment. |
Recent official incidents can explain slow answers, intermittent timeouts, elevated errors, or Claude Code delays even after recovery.
Claude API (api.anthropic.com)
No component listed
No component listed
claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork
claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork
claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork
claude.ai, Claude API (api.anthropic.com), Claude Code, Claude Cowork
claude.ai, Claude API (api.anthropic.com), Claude Code, Claude Cowork
claude.ai, Claude API (api.anthropic.com), Claude Code, Claude Cowork
When Claude still works but feels slow, separate service-side latency from local browser, network, account, prompt, or tool-specific causes.
Claude may be slow because of degraded service, high demand, a recent incident, model-specific latency, account limits, browser issues, network routing, or Claude Code/API problems.
No. Claude can be slow while still partially working. A full outage usually means Claude.ai, the API, or Claude Code is unavailable or returning repeated failures.
Claude Code can be slow because of API latency, model congestion, long context windows, tool calls, CLI or IDE state, authentication, rate limits, local network issues, or a broader Claude incident.
The official status page can be green while a single model, region, account, long prompt, browser session, API route, or Claude Code workflow is still slow. Check recent incidents, component status, and whether the delay happens across web, API, and Claude Code.
If Claude Code feels slower, first separate service latency from workflow size. Large repositories, long context, many tool calls, model changes, usage limits, and local IDE state can all feel like degradation even without a full outage.
If Claude is getting slower over several sessions, compare recent incidents with model choice, long conversations, prompt size, uploaded files, browser state, account limits, and whether Claude Code or API requests show the same slowdown.
Claude can take long to respond when the model is under load, the prompt is large, the conversation has too much context, tools or file uploads are involved, API requests are timing out, or a recent incident is still recovering.
Yes. Usage limits, quota pressure, rate limits, or reduced model availability can make Claude Code and API workflows feel slower or less reliable even when Claude is not fully down.
Terminal-specific slowness can come from PowerShell, antivirus scanning, Node/runtime issues, long sessions, MCP tools, large context, or a local shell setup. Compare another terminal, a smaller project, and a fresh Claude Code session before assuming a Claude outage.
Check the Claude API component and recent incidents. API latency can affect Claude Code, developer apps, and long-running requests even when Claude.ai still appears to work.
Yes. Claude.ai, Claude API, Claude Code, and Console can have separate latency or error patterns, so check the component that matches your workflow.
Check recent incidents, retry after a short wait, test another browser or network, reduce large prompts or file uploads, and compare whether the issue appears on web, API, or Claude Code.