Is Atlassian Rovo down?

Rovo is experiencing reduced reliability, with streamed message responses failing for some requests. This is affecting the Rovo product and related internal services that depend on the same path. Affected individuals may see errors, failed responses, or delayed replies when using Rovo Chat. Our teams are investigating the issue and more information will be shared within the next 1 hour.

Rovo continues to experience degraded performance and this also affects connected Rovo experiences such as agents and other features that rely on the same request path. Affected users may see errors, failed responses, delayed replies, or permission-related failures. Our teams have identified the root cause of this incident and is now actively working on mitigating the issue. We will provide further updates within an hour or sooner if we have any significant progress to share.

Our teams have taken corrective steps to mitigate the issue and are observing some improvements. We continue to monitor the situation and will take necessary actions until the incident is fully recovered. We will provide further updates within an hour, or sooner if we have significant progress to share.

Our teams continue to work on recovery until the incident is fully resolved. We will provide further updates within an hour, or sooner if we have significant progress to share.

We have now implemented a fix and services have been restored to normal operation for affected customers. On August 18 2026, users experienced reduced reliability, with streamed message responses failing for some requests. This also affected the Rovo product and related internal services that depend on the same path. Affected users also observed errors, failed responses, or delayed replies when using Rovo Chat.

**SUMMARY** On August 18, 2026 between 06:16 and 10:48 UTC, Rovo Chat and connected AI experiences suffered significant degradation. A configuration change inadvertently activated an experimental, resource-intensive AI capability across a much larger share of traffic than intended. This triggered a cascading failure in our shared caching infrastructure, leading to widespread request failures across multiple AI-powered surfaces. **IMPACT** During the impact period, customers experienced failed or unresponsive interactions when sending messages to Rovo Chat and Agents, as well as issues with select AI features across Jira and JSM. Requests either failed outright or returned intermittent errors or access-denied responses. **ROOT CAUSE** A configuration cleanup update contained an unintended change that inadvertently enabled a new Atlassian experimental agent - Marathon Agents - designed for limited and controlled rollout, for nearly all incoming requests. You can learn more about the rollout of Marathon Agents here: [https://community.atlassian.com/forums/Rovo-articles/What-s-New-in-Rovo-Agents-Your-Bi-Monthly-Update/ba-p/3242351](https://community.atlassian.com/forums/Rovo-articles/What-s-New-in-Rovo-Agents-Your-Bi-Monthly-Update/ba-p/3242351). This capability places significantly higher demand on our shared caching infrastructure than standard workloads. The sudden load on the caching infra led to a series of cascading failures in our systems which took some time to recover. **REMEDIAL ACTION PLAN & NEXT STEPS** The incident was resolved by rolling back the configuration change, disabling the experimental capability for all users, and scaling up infrastructure to absorb residual load. To help prevent recurrence, we are taking the following actions: - **Cache isolation:** Introducing dedicated, isolated caching instances for critical operational flows (such as permission checks and session state) and high-intensity AI features, to limit blast radius from any single component - **Improved pre-production validation:** Strengthening our staged rollout and soak-testing processes to catch resource-intensive changes before they reach production at scale - **Architectural hardening:** Reviewing the design of resource-intensive AI capabilities to ensure they are resilient to infrastructure pressure and safe for broader rollout We sincerely apologize for the disruption this caused. We are committed to the reliability of Rovo and our AI experiences, and these improvements will be prioritized accordingly. Thanks, Atlassian Customer Support