Enterprise AI adoption is accelerating, but so is the operational risk that comes with autonomous agents operating across cloud environments. For MSPs in particular, the challenge is not just deploying AI safely for one customer, but doing it consistently across many AWS environments, each with different data, identities, and risk profiles.
At AWS re:Invent 2025, Rubrik introduced
Rubrik Agent Cloud for Amazon Bedrock AgentCore to address that gap. The integration is designed to give security teams and service providers clearer visibility into AI agents, stronger governance controls, and practical recovery options when agents behave in unexpected or harmful ways.
A control layer for AI agents at scale
As AI agents become increasingly embedded in customer workloads, MSPs require a method to manage them without introducing operational complexity. Rubrik positions Agent Cloud as a centralized control point for that work.
David Kosman, Vice President of Global Managed Service Providers at Rubrik, stated that the platform is designed to provide MSPs with a comprehensive view of agent activity across customer environments. “Rubrik Agent Cloud helps Managed Service Providers manage AI agents across their customers’ AWS environments by providing a single place to see and control the entire AI agent process, from start to finish,” he said.
According to Kosman, visibility starts with automatic discovery. Agents built on Amazon Bedrock AgentCore, along with custom agents, are identified and mapped so teams can understand their blast radius across infrastructure. From there, MSPs can continuously monitor activity, enforce behavior and access policies in real time, and retain immutable audit trails that capture data, identity, and application context.
The result, he added, is the ability to automate complex tasks while still managing the expanding attack surface introduced by generative AI.
Governance that moves beyond observation
AI governance often breaks down when teams can see what an agent did but lack the tools to stop or redirect it in the moment. Rubrik Agent Cloud is structured around three core capabilities - Agent Monitor, Agent Govern, and Agent Rewind - to close that gap.
Governance features allow MSPs to define how agents should behave, what they can access, and which actions are permitted. Those controls can be enforced as agents operate, rather than after a problem surfaces. For multi-tenant service providers, this creates a more consistent and repeatable way to apply guardrails across customers without building custom controls for each environment.
Recovery designed for agent-driven errors
One of the most difficult issues with agentic AI is remediation. When an agent modifies data, configurations, or workflows incorrectly, traditional recovery methods are often slow and disruptive.
Kosman pointed to Agent Rewind as a key capability for MSPs managing uptime across multiple tenants. “Agent Rewind is a vital feature for reducing risk and protecting uptime across multiple tenants by enabling the surgical, instant reversal of agent-driven mistakes,” he said.
Integrated with Rubrik Security Cloud, Agent Rewind allows service providers to roll back specific, undesirable actions without triggering full-system restores. Kosman noted that MSPs can undo unwanted changes quickly, avoiding the downtime and data loss typically associated with broader recovery processes, while maintaining continuous protection and immutable recovery for customer environments.
What the AWS Resilience Competency means for partners
Alongside the Agent Cloud announcement, Rubrik also achieved the AWS Resilience Software Competency in the Recovery category. For partners, that designation carries practical implications beyond a badge.
Kosman said the competency validates Rubrik as a foundation for partners building AI-focused resilience and governance offerings. “Rubrik earning the AWS Resilience Competency in the Recovery category validates that our technology is a trusted and reliable foundation for partners to create their own AI-driven resilience and governance services,” he said.
Because the platform integrates data, identity, and application context, MSPs can use Rubrik’s certified solutions to design services aligned with each customer’s availability and recovery requirements. That opens the door to more structured AI governance offerings tied directly to uptime, resilience, and operational assurance.
For MSPs supporting customers on Amazon Bedrock, Rubrik Agent Cloud is positioned as an operational layer that sits between AI innovation and real-world risk. It focuses on three areas that matter at scale: knowing where agents exist and what they touch, enforcing clear rules as agents act, and recovering quickly when automation goes wrong.
As AI agents move from experimentation into production across AWS environments, those capabilities are becoming less optional. Platforms that combine visibility, governance, and recovery are likely to shape how service providers deliver AI safely, predictably, and at scale.
