Cohesity Agent Resilience: Protect and recover the infrastructure that powers AI agents

Today, Cohesity is introducing Cohesity Agent Resilience, a new capability in Cohesity Data Cloud that helps organizations protect agent state and recover AI agents to a trusted version when something goes wrong. 

AI agents are changing enterprise operations. Unlike AI assistants that help people complete tasks, enterprise AI agents execute multi-step workflows without a human reviewing every step—building memory and context with every interaction. That accumulated context makes an agent better at decisions and more effective at operating.

When memory is compromised or configurations drift, recovering the applications and data affected by the agent is not enough. Organizations need a way to restore the trusted state of the agent itself. Without a trusted recovery point, you may have to rebuild an agent and start over.

Cohesity Agent Resilience extends the cyber resilience foundation organizations already trust for data protection into agentic environments—helping AI/ML, security, and IT teams protect agent state, identify when that state becomes compromised, and recover faster when incidents occur. 

It is now available to select customers for agents built on AWS Bedrock.

The part of your agent nobody's protecting today

Traditional applications depend on data and configuration. AI agents introduce a new stateful tier—agent state—the complete snapshot of everything an agent needs to know, remember, and act on at a given moment. It includes:

• Memory, such as conversation history and learned context.
• Configuration, including prompts, personas, and system instructions.
• Guardrails that define expected boundaries.
• Credentials and permissions used to access tools and services.
• Workflow and operational context that influence what happens next.

Most organizations aren't protecting this state today. A prompt injection or an unintended change can leave an agent operating normally on the surface, while acting on information it should never have trusted.

How Cohesity builds resilience for AI agents

Cohesity Agent Resilience addresses this in three ways:

1. View a comprehensive map of the agent and its dependencies.

Figure 1: A unified view of an agent and the state it depends on.

Before protecting an AI agent, organizations need visibility into the environment that supports it. Cohesity Agent Resilience automatically discovers supported agents and maps the stateful components and the dependencies that shape their behavior. Agent topology brings the agent, its memory, configuration, and data resources into a unified Application Group in Cohesity Data Cloud. 

By establishing a single resilience boundary around the agent and its critical dependencies, teams gain visibility into what must be protected and what could be impacted.

2. Protect the state that drives agent behavior.

From there, Cohesity continuously protects the core pieces that determine how an agent behaves, establishing trusted recovery points before incidents occur.

3. Recover the agent’s trusted state.

Figure 2: The recovery workflow allows administrators to select specific agent memories and operational context from a known-good recovery point.

AI control towers and observability platforms can detect suspicious behavior, but an alert that something's wrong is only the beginning—you need a way to act on it. Cohesity Agent Resilience enables precise point-in-time recovery of agent memory, configuration, and runtimes. Teams can pinpoint when suspicious behavior began, select a recovery point from before the incident, and restore only the affected state. 

Preserve knowledge, restore trust

As agents become embedded in critical workflows, their memory and configuration become part of the operational environment organizations must protect. Cohesity Agent Resilience helps organizations:

• Establish protection before an incident occurs.
• Recover affected state without rebuilding the entire agent.
• Deploy enterprise agents with greater confidence.

The outcome is not simply another backup of agent data. It is faster recovery from agent failures and attacks, reduced risk and compliance exposure, and the confidence to scale agentic AI. 

Get started with Cohesity Agent Resilience

Agents are already making decisions inside your business. The question isn't whether to trust them—it's whether you can recover when they get it wrong.

Cohesity Agent Resilience gives you that path back, starting with agents built on AWS Bedrock.

Resilience means the ability to restore trust, not just detect risk. Contact your Cohesity account team to get started. 

More resources:

• Learn more about Cohesity Data Cloud
• Read the press release