Is Codex or Claude down? How to check official status pages

Quick answer: for Codex, check status.openai.com and look at the Codex group, which covers Codex Web, Codex API, CLI and VS Code extension. For Claude and Claude Code, check status.claude.com. The old status.anthropic.com address now redirects there. Both pages offer feeds and a machine-readable status endpoint, so you can get alerts instead of refreshing.

Key facts

• OpenAI: status.openai.com. The Codex group covers Codex Web, Codex API, CLI and VS Code extension. Codex in ChatGPT Desktop is listed under ChatGPT.
• Anthropic: status.claude.com. Components include claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork and Claude for Government.
• Feeds: status.openai.com/history.rss and status.claude.com/history.rss (Atom versions too).
• JSON: /api/v2/status.json on both hosts returns an overall indicator, such as none or minor.
• A slow model isn't always an outage. Check the incident title for the model name.

Codex and ChatGPT: status.openai.com

OpenAI's status page groups services into APIs, ChatGPT, Codex, FedRAMP and Ads Platform. The Codex group has four components: Codex Web, Codex API, CLI and VS Code extension. If you use Codex inside the ChatGPT desktop app, also watch Codex in ChatGPT Desktop and ChatGPT Work, which sit in the ChatGPT group. Each incident lists what it affects. On 13 October, for example, the page showed "Degraded performance in GPT-6.1 Sol Codex and Work mode" affecting ChatGPT and Codex.

To get alerts, use the page's Subscribe to updates button or the RSS feed at status.openai.com/history.rss.

Claude and Claude Code: status.claude.com

Anthropic's status page lists claude.ai, Claude Console (platform.claude.com), Claude API (api.anthropic.com), Claude Code, Claude Cowork and Claude for Government as separate components. If Claude Code fails but claude.ai works, check the Claude Code and Claude API rows. You can subscribe by email, SMS, Slack, Microsoft Teams, webhook or RSS, and the feeds are at status.claude.com/history.rss and history.atom.

How to read an incident

Each incident has a status that moves through stages, such as investigating, identified, monitoring and resolved. Monitoring means a fix is in and the provider is watching recovery, so things may still be a bit flaky. Component states range from operational through degraded performance and partial outage to major outage. Read the incident title too. Many incidents are limited to one model, for example "Elevated errors for Claude Opus 5.5" in Anthropic's recent history, so switching models can be a quick workaround.

For scripts and dashboards

Both hosts answer at /api/v2/status.json with an overall indicator and a short description. On 13 October, for example, both returned a "minor" indicator. Both also serve /api/v2/summary.json with per-component states. Polling every few minutes is plenty. These endpoints are public and need no key.

Is it an outage or just your limits?

• Codex: run /status in the CLI to see your remaining limits, or open the usage dashboard.
• Claude Code: run /usage for plan limits. /status shows version, model, account and connectivity, per the commands reference.
• Claude apps: Settings > Usage shows your next reset time.
• If the status page is green and you're within limits, check your own network or proxy before assuming an outage.