Komodor has launched its Agentic Operations Platform, adding ready-to-run AI workflows for production operations and a shared control layer for organizations that build or import their own agents. The platform is available now, the company said, and is aimed at teams managing incident response, delivery health and infrastructure cost in production environments.
The recent launch is about operational automation versus general-purpose AI. Komodor said the platform combines its existing AI SRE infrastructure with workflows for troubleshooting and incident management, alert intelligence, proactive reliability work, cloud and Kubernetes cost optimization, and software-delivery tasks such as change-risk control and CI/CD health remediation.
For an operations team, the practical change is a packaged way to deploy several specialist agents and integrations as an end-to-end workflow instead of assembling each automation independently. Komodor said its solution workflows include more than 50 prebuilt agents, skills, integrations and MCPs. Teams can change steps and routing or add their own agents, according to the release.
The platform also addresses a problem that becomes more consequential when agents can reach production systems: a successful prototype does not establish that an agent has the right context, permissions, auditability or cost controls to operate safely. Komodor said custom agents can use shared organizational context, persistent memory, knowledge graphs and operating controls, whether they are created with the company’s SDK, converted from an existing runbook or imported from third-party frameworks.
Its stated governance model includes role-based policies for agent invocation, credentials and tools; guardrails that inspect inputs, tool calls and model responses; human approval gates for risky actions; spending limits; and an audit trail. Those controls are relevant to platform and SRE teams that want to use automation for repetitive production work without assigning every proposed action the same level of access or autonomy.
Komodor also said customers can shadow-test new agent versions, compare performance against validated scenarios and route tasks to an appropriate model before promoting changes. The release does not specify which tests or production outcomes qualify as validation, how the controls integrate with particular identity systems, or which actions require human approval by default. Organizations evaluating the platform will need to confirm those details against their own change-management, security and incident-response policies.
The company is positioning the platform around a growing operational burden from faster software delivery. More frequent application changes can add alerts, support requests, maintenance work and cloud costs, while operations teams still need to establish whether an issue requires investigation, remediation or escalation. In that setting, an agentic workflow can be useful only if it has access to relevant signals and a defined route from analysis to action.
Komodor’s earlier platform focused on cloud-native and Kubernetes operations, and the new offering extends that foundation to customer-built agents. The company said agents can run on-premises, in a cloud environment or through Komodor Cloud, using different models and providers. It also said the platform can turn an existing skill, script or runbook into a governed agent, although the announcement does not detail supported runtime configurations or deployment requirements.
For IT operations leaders, the launch offers an automation framework that joins AI SRE, delivery operations and cost-management workflows under one set of controls. Its potential value is reducing the manual effort required to connect alerts, operational context and approved actions. That value will depend on how accurately the workflows map to an organization’s infrastructure, permissions and escalation practices, rather than on autonomous execution alone.