[India] - Degraded Performance in Agents Evaluations

Update timeline

1. monitoring Jul 18, 2026, 08:03 AM UTC

   We have identified the cause of the degraded performance impacting Evalutions for Agents in India region and we have applied a mitigation and are monitoring the situation. Impact: Users may experience timeouts when running evaluations. Additional updates will be provided We have identified the cause of the degraded performance affecting evaluations for agents in the India region. We have implemented mitigation measures and are monitoring the situation. Impact: Users may experience timeouts when running evaluations. We will provide additional updates as we work toward a resolution.as we move toward resolution.

2. resolved Jul 18, 2026, 08:30 AM UTC

   We have identified and resolved the cause of the degraded performance affecting Agent Evaluations in the India region. Impact: Users may have experienced timeouts when running evaluations. This incident is now resolved. No further updates will be provided.

3. postmortem Jul 29, 2026, 11:43 AM UTC

   ## Customer impact On July 18, 2026, customers in the India region may have experienced degraded performance, timeouts, or failures when using Agent Evaluations and Health Score. Some published Conversational Agents may also have been unable to initialize conversations. Serverless Robot jobs may have started or executed more slowly than normal. ## Impact timeline All times below are on July 18, 2026, in UTC. | Time | Affected services | Customer impact | | --- | --- | --- | | Approximately 07:00–07:09 | Agent Builder | Customers may have experienced degraded performance, timeouts, or failures in Agent Evaluations and Health Score. Some published Conversational Agents may also have been unable to initialize conversations. | | 07:09–08:30 | Agent Builder and Serverless Robots | The Agent Builder impact continued. Serverless Robot jobs may also have started or executed more slowly than normal. | | 08:30–09:06 | Serverless Robots | Agent Builder service had recovered. Serverless Robot jobs may still have started or executed more slowly than normal until service was fully restored. | ## Root cause An unusually large traffic spike in the India region exceeded the available service capacity for Agent Builder and Serverless Robots. ## Detection The issue was detected through automated service alerts. Our engineering teams began investigating and published customer notifications for the affected Agent Builder and Serverless Robot services. ## Response We restored service by increasing regional capacity, controlling the abnormal workload, and monitoring the affected features until performance stabilized. ## Follow up We are improving capacity safeguards, workload controls, isolation, and the performance of affected request paths to reduce the likelihood and customer impact of similar traffic spikes.