Komodor unveils Klaudia AI multi-agent SRE platform

Komodor has released a new extensibility framework that turns its Klaudia AI product into a multi-agent platform for site reliability engineering in cloud-native environments.

The framework coordinates multiple specialised AI agents that investigate incidents in parallel and carry out remediation across infrastructure layers. It targets operational work where faults can span Kubernetes clusters, cloud services, GPUs, networking and storage.

Komodor plans to preview the technology at KubeCon Europe in Amsterdam next week and demonstrate the new multi-agent capabilities.

Multi-agent design

Komodor described the framework as a modular system with several agent types, including workflow agents that handle detection, investigation and remediation, and subject matter expert agents focused on domains such as Kubernetes, AWS services, GPUs or deployment tools.

The platform includes more than 50 specialised agents, and customers can add their own agents and tools to run alongside Komodor's agents within the same workflow.

The approach mirrors how site reliability engineering teams often work during outages, splitting tasks across domains. Different specialists check application behaviour, cluster health, network paths, storage performance and external dependencies. Komodor's orchestration layer coordinates agents covering each area and combines their findings into a single investigation.

"Most AI tools for operations focus on summarizing telemetry rather than resolving incidents, but complex outages require specialists from multiple domains working together to understand what's happening across the stack," said Itiel Shwartz, Co-Founder and CTO, Komodor.

Extensibility options

Organisations can connect existing services and tools through MCP or an OpenAPI specification, enabling teams to keep their established operational systems while applying the multi-agent approach.

Komodor said the orchestration layer can call on different agents during an investigation based on what it discovers. The structure is intended to reduce reliance on a single general assistant processing large volumes of data at once and to keep each step tied to the relevant operational context.

"The Komodor platform's new extensible architecture replicates this collaborative process using specialized agents that encode operational knowledge and work together to diagnose and resolve issues," said Shwartz.

Komodor said early adopters have used the framework to build custom agents tailored to their environments and processes. Examples included agents that cross-reference CI/CD pipelines to link failures with recent changes, integrate with database management tools to check query or connection-pool issues, and search past incident channels for similar symptoms and prior resolutions.

Product scope

Komodor's announcement follows wider adoption of AI tools in operations teams, where much of the market has focused on analysing and summarising logs, traces and metrics. Komodor is pitching a system that extends into coordinated investigation and remediation across a broader range of cloud-native components.

Komodor described the launch as part of its evolution from automated troubleshooting to what it calls an autonomous AI SRE platform. It framed the shift around extensibility, with customers able to bring domain tools and internal knowledge into the same process as Komodor's pre-built agents.

The multi-agent framework for Klaudia AI is available immediately through Komodor and its business partners worldwide. Komodor also plans to demonstrate the new capabilities at KubeCon Europe in Amsterdam.