Komodor adds memory layer to Klaudia for recurring incident response

Komodor Ltd. has released Klaudia Memory, a new capability for its autonomous site reliability engineering platform that the company says lets its AI agents retain context from earlier incident investigations. The feature is available immediately and is aimed at production teams dealing with distributed cloud-native systems, where recurring failures can be hard to trace across changing infrastructure.

The company did not disclose pricing, customer adoption figures, revenue impact or performance benchmarks for the release. Komodor is positioning the update as a way to move beyond generic AI copilots in operations work, where the useful context is often specific to one company’s architecture, dependencies and operational history.

Klaudia is Komodor’s AI-native troubleshooting platform for SRE and DevOps teams. With the new memory function, Komodor says its agents can use the results of previous root-cause investigations when responding to new incidents. That could mean recognizing a pattern such as a memory leak following a recurring batch job, or connecting a familiar timeout to the same underlying component.

The pitch is straightforward: if an AI troubleshooting system can remember what was already ruled out, it can avoid repeating parts of the investigation. Komodor says Klaudia Memory can also learn which signals in a given environment tend to be harmless, reducing unnecessary alerts and helping teams preserve operational knowledge that usually sits with senior engineers.

Komodor co-founder and Chief Technology Officer Itiel Schwartz said the company sees generic AI copilots as under-equipped for enterprise production operations because each organization has its own systems, constraints and incident history. According to Schwartz, Klaudia Memory is designed to make that institutional knowledge available during investigations without training on customer data or mixing information between customers.

AI ops tools are chasing context

The release lands in a crowded part of the software market where vendors are trying to apply generative AI to incident response, observability and infrastructure operations. Coding assistants have shown clearer adoption because codebases and developer workflows are easier to package into repeatable products. Production operations are messier: the same alert can have different meanings depending on deployment patterns, service ownership, cloud architecture and business priorities.

Komodor’s argument is that SRE automation needs a memory of the customer environment, rather than only broad model training. That is a credible direction for the category, though the company has not provided independent data showing how much faster Klaudia resolves incidents with the memory feature enabled.

Komodor also introduced Headless Klaudia, which lets engineers interact with the AI system from tools outside the main Klaudia interface. The company said teams can trigger investigations from Slack and Microsoft Teams, as well as development environments including VS Code, Claude Code and Cursor.

The company said the headless approach also connects with existing version control and GitOps workflows. Komodor added new application programming interface and mode context protocol integrations so Klaudia can be embedded in internal tools.

Komodor is also expanding Klaudia’s specialist agents beyond Kubernetes. The company said the platform will support additional workloads running on Amazon EC2 and ECS, a sign that it is trying to broaden from Kubernetes troubleshooting into a wider production-operations product.

This story draws on original reporting from SiliconANGLE.