AI Agent is responding normally again. Total disruption: about 1h 5m.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

Resolved. Prompts and notebook runs are back to normal speed. The limit was our own capacity rather than a model provider, and we have since expanded it and moved to faster hardware. Nothing was lost.

The cause is our own capacity, not a model provider: today's traffic has pushed past what we can run at once, so work is queueing before it starts. Runs still complete, but the slowest can take several minutes. Switching models will not help. Signing in and your saved notebooks are unaffected.

Prompts and notebook runs are slower than usual and some are timing out. Signing in and your saved notebooks are unaffected. We're investigating.

AI Agent is responding normally again. Total disruption: about 30 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Impact escalated — AI Agent is now unreachable.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 10 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 10 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

API is responding normally again. Total disruption: about 10 min.

API is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that API is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 15 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 25 min.

AI Agent is responding again. Monitoring to confirm full recovery.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 10 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 15 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

Resolved, along with the API incident it shared a cause with. Runs and run state are back to normal. If a run failed during the disruption, try it again.

Recovering along with the API. Runs are going through normally again and run state is saving. We're watching a while longer.

Same cause as the current API incident. Runs may be slower or occasionally fail, and saving run state may error out. Retrying usually works; switching models will not help.

Automated monitoring detected degraded AI Agent performance. We're investigating.

Resolved. A saved notebook state could fail to load into a fresh session and then block that notebook from running, and retrying replayed the same state. Sessions now start cleanly instead. Your work was never at risk — if a notebook still looks stuck, run any cell once.

Recovering. Save and restore errors have come down and the API is responding normally again. We're watching a while longer before calling it resolved. If a run failed to save its state, try it again.

Server issues on our side are causing errors when we save or restore your notebook state, and idle sessions may be slow to come back. Signing in and your saved notebooks are unaffected, and retrying usually works.

We're seeing intermittent errors and slower-than-usual responses from the API and are investigating.

Kimi K3 is back and selectable again for everyone whose plan includes it. Qwen 3.8 Max continues to run normally. Thanks for your patience through a long one.

Closing this incident. Qwen 3.8 Max has been running normally for days. Kimi K3 is still unavailable, so we've temporarily removed it from the model picker until it genuinely works, and we'll announce it here when it returns. Every other model, including Auto, is operating normally.

Sorry for the silence here. Qwen 3.8 Max is fully restored and running normally. Kimi K3 is still unavailable while we work through a capacity constraint with the provider, and it stays unselectable until it genuinely works. Every other model, including Auto, is operating normally.

Recovery is taking longer than expected. Kimi K3 remains unavailable and Qwen 3.8 Max remains slow while upstream capacity stays constrained. All other models, including Auto, are operating normally.

Kimi K3 is unavailable due to a server issue at our cloud provider, and we expect to restore it within 24 hours. Qwen 3.8 Max is available but currently very slow. All other models, including Auto, are operating normally.

We've deployed a fix on our side. What remains is upstream: Qwen 3.8 and Kimi K3 are still being rate limited, so they may be slower than usual or occasionally fail. All other models are operating normally.

Qwen 3.8 and Kimi K3 are degraded and may be temporarily unavailable — requests to them are being rate limited. Other models are unaffected, so switching is the fastest workaround.

Database & Auth is responding normally again. Total disruption: about 10 min.

Database & Auth is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that Database & Auth is unreachable. We're investigating.

AI Agent is responding normally again. Total disruption: about 10 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

AI Agent is responding normally again. Total disruption: about 10 min.

AI Agent is responding again. Monitoring to confirm full recovery.

Automated monitoring detected that AI Agent is degraded. We're investigating.

Resolved. Qwen 3.8 Max is available again and responding normally. If you switched to another model, you're welcome to switch back.

Qwen 3.8 Max is temporarily unavailable while we restore upstream capacity, and we expect it back within about 30 minutes. All other models, including Auto, are operating normally.

Resolved. Clusy is back online — the app loads, sign-in works, and notebooks run normally again. Sorry for the disruption.

Clusy is currently down. You may be unable to load the app, sign in to your workspace, or run notebooks. We're investigating and will post updates here.

Resolved. The cause was a bug in our real-time messaging layer rather than the AI model: a message sent from the home screen just after the workspace opened could be dropped before processing. If a message of yours went unanswered, please resend it.

We've identified degraded performance in the underlying AI model that began about four hours ago. Sending a message from the home screen may be slow or occasionally fail — please retry if it doesn't go through.

We're seeing degraded AI model performance that may cause slower responses and errors when sending a message from the home screen. We're investigating.

Service has been restored. The API and AI Agent are responding normally again. Total disruption: about 42 minutes.

We detected that the API and AI Agent were unreachable and are investigating.