AI & ML API Status Monitoring

PulsAPI tracks the AI and ML platforms in production stacks, including OpenAI, Anthropic, Hugging Face, Cohere, Mistral AI, Groq, Replicate, ElevenLabs, and Perplexity, plus vector databases like Pinecone. Each is polled every 60 seconds.

When a model provider degrades, your AI features fail even though your own servers are healthy. Treating AI APIs as a monitored dependency lets you detect the outage, attribute it correctly, and fall back gracefully instead of debugging your own code.

Yes. Subscribe to any AI provider and route alerts to Slack, Discord, Microsoft Teams, PagerDuty, or email. Alert rules let you page on a major outage while sending degraded-performance notices to a quieter channel.

It depends which layer failed, and they fail independently. A model API (OpenAI, Claude, Amazon Bedrock) returning 500s breaks every app built on it. A GPU host (RunPod, Replicate, Modal) can be down while model APIs are fine. And an AI product (Copilot, Devin, Cursor, Perplexity) can be broken while the model provider underneath it is healthy, usually because of its own backend rather than the model. The grouped list on this page shows all three layers at once so you can tell which one moved.

In practice the gateway and orchestration layer, not the model. Model APIs are heavily redundant; the failures teams actually hit are rate limits, regional capacity on GPU hosts, and outages in the routing layer between the app and the model. That is why PulsAPI tracks gateways such as OpenRouter and LiteLLM separately from the model providers they route to.