Service status catalog
Is it down, or is it you?
Live status of 99 services that AI products and their agents depend on. Each one is read from the provider's own status page, once for everyone, and the most used APIs are also checked by an independent UpButler probe. Every service has a small JSON verdict an agent can read before it retries.
10 services reporting problems right now
- OpenAI · DegradedDelayed Costs data in the Compliance API
- Anthropic (Claude) · Partial outageElevated errors on platform.claude.com
- Cloudflare · DegradedElevated Errors in China Network Colos
- MongoDB Atlas · Partial outageImpaired Cluster operations in AWS me-central-1 and AWS me-south-1
- Algolia · DegradedIncident on cluster r20-br: We are encountering a major issue that impact all API calls.
- GitLab · Partial outageErrors and slow performance on merge requests in the specific groups.
- Zapier · DegradedDegraded performance
- Twilio · DegradedSMS Delivery Receipt Delays from a Subset of Twilio Short Codes to Multiple Networks in New Zealand
- Slack · DegradedSlack for iOS closing unexpectedly when opening channels
- Clerk · DegradedDevelopment instances only - Apple rejecting emails
AI models and tooling
- OpenAIDegraded
- Anthropic (Claude)Partial outage
- OpenRouterOperational
- Google Gemini & Vertex AIOperational
- Mistral AIOperational
- GroqOperational
- Together AIOperational
- xAI (Grok)Operational
- DeepSeekOperational
- CohereOperational
- PerplexityOperational
- Fireworks AIOperational
- CerebrasOperational
- Hugging FaceOperational
- ReplicateOperational
- ElevenLabsOperational
- DeepgramOperational
- AssemblyAIOperational
- CursorOperational
Cloud and edge
Hosting and compute
Databases and data
Code, packages and workflows
Payments and commerce
Messaging and realtime
Search and web data
Built to be read by agents
Every service has /status/<service>/status.agent.json: under 1 KB, a verdict (proceed, retry, fallback, pause) and how long to wait. The whole catalog is at /status.json. No key, CORS open, ETags for cheap polling.
In an UpButler workspace, declare what your project depends on and an outage of theirs is tagged "likely upstream" on your incident, so nobody, human or agent, debugs your code for someone else's outage.
curl -s https://upbutler.com/status/anthropic/status.agent.json{ "mcpServers": { "upbutler": { "url": "https://upbutler.com/mcp/public" } } }
// then: upstream_status {"service": "openai"}UpButler is not affiliated with the services listed. Status and incident text belong to each provider and are read from their public status sources.