Rafay Systems Introduces a Managed MCP Server Offering to Bring Governed AI Assistance to Infrastructure Operations

New offering gives platform and SRE teams a secure, controlled way to connect AI assistants and agentic applications to Rafay operation context, starting with fleet intelligence cost attribution and incident diagnosis.

Rafay Systems, a leader in infrastructure orchestration for AI and cloud-native workloads, introduced a new Model Control Protocol (MCP) Server capability, giving platform, DevOps and site reliability (SRE) teams a governed interface for asking operational questions about their infrastructure without exporting data, translating API responses or building one-off integrations for every tool. Initial workflows focus on fleet intelligence, cost attribution, and incident diagnosis for Kubernetes environments managed through Rafay, with MCP providing a common foundation for future agentic workflows across Rafay-managed compute resources.

Infrastructure teams already use AI assistants, but those assistants usually cannot see the operational fabric: projects, clusters, namespaces, policies, usage, health and cost signals. Engineers become the integration layer, moving between consoles, command lines, spreadsheets and chat windows to reconstruct context that no single system exposes on its own. The Rafay MCP Server changes that model by connecting AI assistants to Rafay’s governed operational context through a standard protocol and the same RBAC and project boundaries teams already use.

Also Read: CIO Influence Interview with Hugo Dozois-Caouette, CTO and Co-founder at MaintainX

Fleet intelligence and cost attribution
The first workflow helps teams turn fragmented fleet data into answers they can act on. A platform engineer can ask an MCP-compatible assistant to inventory clusters across projects, summarize health and configuration status, identify stale or underutilized environments, and correlate workload placement with node specifications and infrastructure cost. Instead of manually joining cloud bills, Kubernetes state and Rafay inventory, teams can ask for outcomes such as which clusters have the highest cost per workload or which environments are candidates for consolidation.

Incident diagnosis and operational triage
The second workflow applies the same approach to incident response. An engineer can ask which clusters or namespaces are degraded, drill into unhealthy resources, and receive a structured diagnosis for common issues such as image-pull errors, unschedulable pods, crash loops, failed deployments, services with no endpoints, missing configuration and unbound storage. The assistant surfaces likely causes and recommends next steps, while the engineer remains in control of any change.

A foundation for agentic operations
Beyond these two workflow use cases, the Rafay MCP Server gives customers and partners a foundation for building agentic operations experiences on top of the Rafay Platform. Teams can connect MCP-compatible assistants or multi-agent applications to Rafay context without writing bespoke translation layers for every API, tool, or data source. That makes MCP valuable beyond reporting and troubleshooting: it becomes a standardized way to let AI systems reason over infrastructure context while respecting enterprise controls.

“The value of MCP is not that an assistant can answer a question faster. The value is that the assistant can finally operate with the right infrastructure context, inside the same governance model the customer already trusts,” said Haseeb Budhani, CEO and co-founder of Rafay Systems. “With the Rafay MCP Server, teams can ask where infrastructure is efficient, what is unhealthy, or how to build new agentic workflows on Rafay without turning senior engineers into the manual glue between consoles, APIs, and spreadsheets. The protocol is table stakes. The operating model unlocks the potential.”