---
title: "Connected Cloud Resilience: Turning Signals into Protection and Recovery  | SHIFT 2026 On-Demand"
type: "WebPage"
language: "en-US"
url: "https://www.commvault.com/resources/video/turning-signals-into-protection-and-recovery"
date: "2026-11-16T13:44:29-05:00"
modified: "2026-11-16T13:44:29-05:00"
description: "See how shared context, MCP servers, and reusable skills connect cloud security with data protection for faster, repeatable resilience workflows."
image: "https://www.commvault.com/wp-content/uploads/2026/11/Connected-Cloud-Resilience-Turning-Signals-into-Protection-and-Recovery-1.png"
breadcrumbs:
  - name: "Commvault Home"
    url: "https://www.commvault.com/"
  - name: "Resources"
    url: "https://www.commvault.com/resources"
  - name: "Video"
    url: "https://www.commvault.com/resources/video"
  - name: "Turning signals into protection and recovery"
---

# Turning Signals into Protection and Recovery • SHIFT On-Demand

- [Home](/)
- Videos
- Turning Signals into Protection and Recovery • SHIFT On-Demand

ON-DEMAND SESSION

## Connected Cloud Resilience: Turning Signals into Protection and Recovery

See how shared context, MCP servers, and reusable skills connect cloud security with data protection for faster, repeatable resilience workflows.

[Request a demo](/request-demo)

Play Video      ![Video thumbnail](https://img.youtube.com/vi/s9Z4k1LumPI/maxresdefault.jpg)

---

- [Key Takeaways](#Takeaways)
- [About This Session](#About)
- [Related Resources](#Resources)
- [FAQs](#FAQs)

[Back to all on-demand sessions](/shift#watch-now)

---

## Key Takeaways

**Connect Fragmented Cloud Tools**

Share security, sensitivity, and protection context across the cloud engineering stack.

**Use Context to Act**

Help agents make informed decisions across security and data protection platforms.

**Protect by Sensitivity**

Align backup policies and retention requirements with the sensitivity of cloud data.

**Replace Custom Integrations**

Use MCP interfaces instead of maintaining one-off API connections across tools.

**Standardize with Skills**

Turn multi-step prompts into repeatable workflows with defined rules and checks.

**Keep Approval Gates**

Preview proposed changes and require human approval before applying them.

---

## About This Session

**Cloud resilience improves when security and data protection tools share context instead of operating as disconnected systems.** This session shows how MCP servers can connect cloud security, data protection, cloud infrastructure, and agent workflows. A practical demo combines sensitive data findings with protection status to generate and verify policy based Terraform changes for AWS data.

#### **Build on Shared Context**

Cloud incidents often span several tools. Shared context about risk, sensitivity, recovery points, assets, and protection status helps agents coordinate actions without custom integrations for every connection.

**MCP servers expose platform context and actions through a common interface that supports composable, agent-driven workflows.** Security findings can influence protection policies, recovery timing, and retention. Protection status can show where sensitive assets lack a recoverable copy.

#### **From Prompts to Workflows**

The session first uses separate prompts to classify data, check protection, and generate Terraform. It then converts those steps into a reusable skill with rules, approval gates, verification, and reporting.

**Human approval remains essential when automation proposes changes to critical cloud protection policies.**  The workflow previews Terraform, pauses for authorization, verifies completed changes, and creates a stakeholder-ready report.

#### **Scale Protection Automatically**

Tag-based groups can bring new cloud assets under the correct protection model as environments grow, supporting consistent policy without repeating the full workflow.

CONTINUOUS CLOUD RECOVERY

## CloudRecover

Recovery built for the cloud. Proven at scale. Ready for AI.

[Explore solution](/solutions/cloud-recover)

CONTINUOUS CYBER RECOVERY

## CyberRecover

Clean, isolated recovery after cyber disruption.

[Explore solution](/solutions/cyber-recover)

---

## Related Resources

SOLUTION BRIEF

## AWS Cyber Resilience

Simplify backup management and strengthen resilience for AWS workloads.

[Read the brief about AWS Cyber Resilience](/resources/solution-brief/cyber-resilience-for-aws-workloads-with-clumio)

WHITEPAPER

## Air-Gapped Protection

Preserve cloud data integrity with secure, isolated protection.

[Read the whitepaper about Air-Gapped Protection](/resources/whitepaper/secure-immutable-air-gapped-data-protection-with-clumio)

eBOOK

## Building Cloud Resilience

Discover, protect, recover, and rebuild cloud environments.

[Read the ebook about Building Cloud Resilience](/resources/ebook/building-cloud-resilience)

---

## Frequently Asked Questions

What is connected cloud resilience?

Connected cloud resilience brings security, infrastructure, data protection, and recovery tools together through shared context and coordinated workflows so teams can identify risk, close protection gaps, and recover consistently.

How does MCP support cloud resilience workflows?

MCP servers expose platform context and supported actions through a common interface. Agents can coordinate across tools without requiring a custom point-to-point integration for every workflow.

How can data sensitivity shape protection policies?

Sensitivity findings can determine which assets need protection and how long data should be retained. Policies can differ for financial, personal, and non-sensitive data instead of using one blanket rule.

What is the value of a reusable skill?

A skill packages workflow steps, rules, guardrails, approvals, verification, and reporting into a repeatable process, helping teams deliver more consistent outcomes across environments.

How are automated changes kept under control?

The workflow previews generated Terraform, pauses for human approval, applies authorized changes, verifies the result, and creates a report showing the updated protection posture.

---

## Transcript

View full transcript

Hey, everyone. Thank you for joining us in. This session will be doing a deep dive on cloud resilience.

We’ll specifically cover how you can process and include risk and sensitivity signals in your data protection recovery workflows using the MCP server.

Quick round of introductions before we get started.

I’m Akshay Joshi, I’m a VP of product here at Commvault. And I’m primarily focused on the Clumio suite of products.

I’ll be covering a host of things today. First, I’ll be talking about how cloud operations have evolved over time, with a focus on headless operating models.

We’ll also take a look at why SaaS products are investing heavily in building MCP servers. Next, we’ll explore the cloud engineering stack and how a complex stack necessitates a shared context layer. And we’ll close with how MCB can transform interoperability between the various pieces of that cloud engineering stack and run autonomous workflows.

And I’m joined by Arun and Lawrence today.

I’m Lawrence Chang, Chief Engineering Officer for Clumio and actually his counterpart for development.

I’ll be walking through an example of how ChromeOS and this remote MSPs can bring security and data protection together.

Hi, everyone. I am Arun Noah, VP of product at Waze. And today I’m going to talk about the Waze agents and how we can actually collaborate with other agents in the industry and how that basically help customers bill integration faster and act in the speed of machine.

All right. Before we begin, let’s talk about why we are here. Simply put, fragmented tools lead to fragmented workflows, and these fragments can be across teams, products and processes within a company. And that can significantly slow you down, especially when you’re dealing with a serious incident.

I want to share with you three stats to drive home the point.

96% of the organizations agree that agent success depends on a seamless integration across various platforms. HashiCorp reports that organizations use at least five tools to manage their cloud infrastructure, and will learn more about it when we get to the cloud engineering stack and we examine, what the stack looks like. And finally, 83% of teams use four more tools during a live incident, mainly because the incident context itself can be fragmented across all of these tools.

Right? And mind you, this is during a live incident when when time is most crucial.

Now to build integrated workflows across platforms. These platforms need to operate in a headless manner. Back in the day, engineers would write and maintain custom integrations across platforms using APIs and SDK.

Then came infrastructure as code, just declaring the desired state was enough to manage the infrastructure right.

And most recently, MCP servers expose the platform context and actions to a common interface called the Model Context Protocol, upon which workflows can be built. Now, this evolution all the way from IDE to the comfort of your favorite Lem, has enabled enterprises to focus on composable actions and resulting outcomes rather than building and maintaining integrations.

If you look at this evolution closely with APIs and SDK, or even with infrastructure as code, it was you making decisions still, right? But with MSPs, that’s completely changing. Now, any agent can make all of these decisions on your behalf and run workflows autonomously.

And that brings us to why size companies are investing heavily in MQB servers. Number one, interoperability is huge, right?

We already saw that managing cloud infrastructure spans multiple tools, often more than five, and MXGp server is a gateway to the broader ecosystem because of the shared context across all of these platforms.

Number two is agent autonomy. In order for agents to truly take actions autonomously. We need to provide them with context first. And that’s what the MCP enables, right?

Agents learn how to orchestrate and automate using these tools. And since the context itself is built into the MTP server, we don’t need users that are well-versed in data protection or data security to operate these products. Right? As long as they know how to ask the right questions, they’re good to go.

And finally, API frameworks can be limiting. Imagine the user wants an executive summary on everything that happened in or the past week. Right now, there’s no single API that can give you a custom summary of everything from operations to compliance to consumption. But an alarm can generate one with the help of the context and data that has come to meow.

And of course, the broader LM capabilities right, and I talked about interoperability between SaaS platforms. But the why becomes very clear when you look at this cloud engineering stack. Right. No one is running Lumiere or any product for that matter in a vacuum. Managing your infrastructure requires a whole set of tools ranging from platform engineering to data security and data protection, to observability and monitoring, and shared context across these platforms is absolutely crucial.

Let’s take a very simple example. You’re dealing with a security incident and you’d like to recover your data. You need to know what time you’d like to recover the data too. Basically, what’s the last clean recovery point? The information about that point is in a data security product that you’re using. In this case, let’s say it’s Waze. Now that information has to be shared across the data security and the data protection platforms so that, you know, your recovery workflow can be built based on that point in time.

Right. And the recovery point itself is the shared context across the data security and data protection platforms. In this scenario.

If you zoom in, we can, take a look at a more concrete example. Now, let’s say this identifies sensitive data. Basically whether your data has PII or Phi, right. Now that data drives a lot of your data protection policies within the within a year.

And basically how long you need to retain that data is dictated by the sensitivity of the data itself, right? In case of financial information, you need to be compliant so that the data needs to be retained for seven years in case of your personal health information. HIPAA mandates that the data being for eight years, right. Now these two platforms need to talk to each other in order to exchange that context and run like a well-oiled machine.

Let’s take a look at what that integration look like in a pre MXGp world. First, you need to build custom code for every such connection. With the APIs of both platforms. Right. And you need to keep that code up to date as API signatures change. Which is a very common occurrence.

And you need to maintain that code like every other piece of code in your repository.

You need to manage security and compatibility with the other, pieces of your code. Right. And finally, you need to run that code in some infrastructure somewhere, right? This could be a CI, CD pipeline or a machine within which you can run this code. Right. You need to maintain that infrastructure as well. And finally, you have to do this for every single custom integration that you build across the stack.

Knowing how big the stack is, it can be a tedious undertaking.

Now the MCP server takes care of leveling that playing field across the entire stack with shared context and actions. Now think of all the integrations that you can build across the stack.

The SPM tools dictating how the data gets protected, KSP and threat intelligence tools dictating how and when recoveries happen.

The resilience pipeline itself can be fully autonomous at that point.

And that’s the whole point of an MCP server. It’s a gateway to the broader stack, like I said. And the agents built using these MTP servers can make decisions autonomously based on the shared context across these platforms. The entire stack can be run autonomously with one or more agents, and these agents can manage data protection, recovery drills, and other mundane workflows for you.

On your behalf. And with that, we are introducing the new Amqp server.

All right. Lumion has always followed a headless operating model and average gloomier user logs into the portal once every 37 days.

We’ve had a robust API framework when many other data protection vendors were still hung up on building pretty looking UI.

And we built a Terraform module before it was cool and before all the data protection vendors were even thinking about it. And now we are announcing the availability of the new Amqp server. We followed four key tenets while building the MCP server. The MCB itself is interoperable with the other MCP servers across the stack. It’s non-negotiable. And number two, it’s token efficient.

Tokenomics is a huge part of boardroom discussions lately, and we wanted to build an efficient MCP server that our customers would love using. Number three, it’s secure. It’s human in the loop wherever necessary because we are dealing with crucial data here, right? And finally, it’s not just about the MCP server. It’s about the tooling and the skills that are built on top of the MCP server that makes it more deterministic, right?

We want our customers to be able to build repeatable and predictable workflows using our MCP server, and the skills library is what makes a probabilistic system deterministic in nature.

and with that, I’m going to hand it off to Iran to share his perspective on data security, data protection, and the MTP server.

Thank you. Akshay. Basically, I want to take the opportunity to show a real example of how risk in the cloud exists and their relationship to data risk. Let’s take the latest bit. The shell, vulnerability. It’s a new zero day vulnerability that was revealed that basically every PHP, that is running in your cloud environment and exposed to the internet.

Basically it create attack bus for external attacker to actually do remote code execution with which basically we arm our customers with three agent that can make their life much easier, handle the day to day operations. The first one is the red agent. The red. They just basically acts as your personal sister. He scanned the entire environment and understand which kind of server actually hosted a PHP that are exposed to the internet.

The red agent identify the DOS server and there was application endpoint host a PHP, and then he know basically how to activate the exploitation from the outside the building. He managed to get into the environment. He also understand the underlying technology and the connections we have with the VS graph. And in this specific case, through the exploitation from the outside, it managed to get in and have access to a container.

This container had laterally moved to a database containing PII information. The second agent that come very handy is our blue agents. The blue agent. Basically, keep on monitoring your environment and understand if you have any kind of suspicious activities through getting telemetry from the with sensor and other signals in your cloud environment. The moment the blue agent understood that you are now under attack in this specific case, someone is trying to compromise through the PHP, the shell, the blue agent can actually interact with our third agent, which is the green agent, which is the fixer, and start to activate some kind of like remediation plans in order to stop the attack.

For us, it can actually stop all the processes through the sensor. Second, it actually can activate in meantime, any kind of mitigation like applying WAF rule or even tell the sensor to apply specific rules to block the attack. And third, basically, the moment we understand that we have an ongoing, fix available, it can actually understand who are the right developer that push this code into production, managed to assign them, the issue, offer them to do any kind of remediation related to patch management or basin based images through resource.

And this is how customers are able to remediate fast and interact with all the agents in, in a very a generic way in the speed of machine.

When we talk about any vulnerability when you talk about any zero they obviously unify. And collaborating with the broader ecosystem is something that is extremely important. Now in this specific example with an Lumiere, basically we want to have an integration together. And MCP is the model way for any two companies. And any two products can actually integrate very, very fast.

From the other hand, if basically we want to understand what kind of operation we can do with complex product in this specific base with Domu, we can actually use the MCP in order to query data, understand what kind of an operation it can do, whether it’s like mutation of an S3, data like protection, do any data policy and etc. and from the other hand, if you want to understand which kind of assets contain, data that are exposed to PHP and understand their ongoing, their ongoing risk, basically, Lumiere can query using the Wiz MCP, the green agent, understand what kind of risk we have and what kind of remediation action we can offer.

And MCP basically minimize the time to do integration. And that is the way companies are able to collaborate much faster, bringing bringing basically value for our mutual customer. And I think MCP is just the layer that help everyone also build a genetic virtual, because MCP is basically the layer that connects. But everyone, every company today, they are building agents in order to help customer run things much faster.

And this specific case when Wiz expose MCP server and cobalt expose MCP server, both sides can actually build their agent workflow for the benefit of the customers. So basically, we can blend our green agent in the future with cobalt MCP. Getting more information about what kind of flight back up policy, what are their assets that doesn’t have any big plan in place.

And on the other hand, humans can create agents that can query with MCP and understand the broader risk, factors you have in the environment related to data, what kind of remediation we can offer, and that can also help customers and not just like build integration but also build the agenda workflow in each and each product. With that, I’m handing over to Laurence to show you a cool them about what ways have built together.

Thanks, Arun.

So we’ve been talking a lot about how MSK servers are the building blocks to a giant tech workflows. As such, in this demo, we’re going to show an example of how remote NTP servers from is and come alongside AWS can help bring the worlds of security and data protection together.

So I have in front of me a cloud code session with connections to our NTP servers, and not while you’ll see a lot of cloud in this demo.

Any agent tech harness, whether it’s cloud code, cloud codecs, Cairo, etc. would work. So let’s first take a look at some of the tooling from Wiz through this part to understand the security risks associated with your cloud environment and help identify exploit resources. Misconfigurations find potential vulnerabilities and discover patches of sensitive data amongst the other capabilities of the Wiz platform.

Move along to the tools for the community. MCP now. All right. So coming up brings a different set of capabilities. Check in production of your cloud environment, managing backup policies, planning recovery and even generating Terraform to connect and protect your assets. So each MTP is useful in its own right. But the interesting part is what happens when we bring these MSPs together in the same conversation.

So in this demo, we’re going to start with a couple of simple prompts to highlight some of the capabilities of our MSPs. But eventually we’ll turn this entire conversation into a repeatable workflow where you can inject a harness such as cloud code could make increasingly informed decisions around security and data protection.

All right, so let’s start with an example of some data in S3 supporting an HR application. So we have three S3 buckets one for year 2024, 2025 and 2026.

These S3 buckets will have been provisioned with Terraform. In each bucket, we’ve organized the data into prefixes for payroll stubs, employee records, and benefice brochures. So clearly some of the data in these buckets will be sensitive, and some others, maybe not so much.

So if you were to click into, let’s say, pay stubs, we’d find data of a financial nature. Information such as bank accounts, deposit amounts and so forth. If we were to click into employee records, we’d find data of a more personal nature. Information such as Social security numbers, data, births, addresses, and so forth. And lastly, with the benefits brochures, we find data around health retirement plans.

Information would be a little bit more comfortable sharing broadly. So not as sensitive per se. So given all of this let’s start off with a simple prompt and see what this has to say about the sensitivity of all this data. So I’ll put the product prompt into cloud.

So right off the bat cloud is going to call into the West MXGp to check on the data within the three buckets. Give this a moment to finish. All right. Which has found the sensitive data in all the buckets. Specifically, a data in the payroll and employee record prefixes have been categorized as financial NPI data, respectively. It’s even broken out the types of financial and personal data that it’s found.

So even here it’s mentioned about bank account numbers, Social Security numbers, etc.. So now that we have the context about the data inside the buckets, the next question we want to ask is is any of this protected with colonial. So again plot is going to call into the Columbia at this time and check on data protection status of these three buckets.

So we’ll wait for this to finish.

Okay. And here’s the problem. None of the three buckets is protected by Columbia. No protection group membership, no backup history. So if someone were to accidentally delete this data or we got hit by ransomware, let’s say you don’t currently have a backup based recovery path for it. So let’s see what it would take to fix this. We want to generate Clumio Terraform configuration to bring these sensitive assets under our protection.

So again, I’ll just going to call it back into the MCP to understand how to generate the appropriate Terraform configuration. What is also going to check on the original Terraform that was used to provision these S3 buckets to begin with. For additional context. I’ll give that a moment to finish up.

All right. So the MXGp has given us a great starting point Terraform for our protection policy. Our protection group, which is essentially a grouping of the buckets together into one entity, and the assignment that associates the group with the policy for protection. And just to note, nothing has been applied yet through cloud we’ve only created the Terraform configuration.

But the colonial MCP always previewed any code before it is applied. And that’s a great thing because if I look a little bit closer, the initial configuration is applying a common retention period across all of our sensitive data, which isn’t quite what we wanted. So let’s refine our approach. We have some retention rules in our mind, and we want to use seven years for financial data, one year for personal employee records and 30 days for non sensitive brochures.

Again, we’re going to call back into the MCP to determine how the configuration should be refined to reflect these rules. A given moment for that to finish.

All right. The Terraform configuration has been updated. Instead of one blanket policy for one protection group, we now have three protection groups that protect along the lines of prefixes across our buckets. So for example, the financial protection group is covering the paste up prefixes across all three buckets. And its policy the financial policy is going to retain its data for seven years.

Likewise, the PII protection group is going to cover the employer records and keep its data for one year and non-sensitive data we kept for 30 days. All right. So that’s a much better outcome than what we had initially. But the one thing is that it took us four prompts to get here. We still had to steer the conversation to get what we wanted.

We had to ask separate questions. Check the response and refine the policies ourselves. It feels like we could do a little bit better. And that brings us to the next part.

All right. So what we did was okay once. But what if multiple teams or myself or even you have to do the same task every day? Maybe every week? We’d rather define the steps and guardrails once so people get the same process every time. So let’s take a new approach. We’ll give a new prompt, inspect the security and protection posture of my application with Wiz and Columbia.

And this time that’s the whole request. And the reason for that is because behind the scenes, the team has created a skill, a skill that’s going to tell Claude how to carry out our intended workflow and knows how to gather findings from ways through the Wiz MCP, check KML protection through the Columbia MCP, and ask AWS for information as needed through the AWS MCP.

And those are the types of retention rules that I want. We’ll generate the Terraform appropriately and also make sure it of course to ask for approval before changing anything. So I’ll wait for that to finish.

All right. So here’s the new plan. This time around, the skill is proposing a Terraform configuration that mirrors what we ended up with on our earlier attempt. Prior to using the skill three protection policies, three protection groups, all aligned to the sensitivity of the data. It’s also gone one step further. So if you look at the second line within this table, it’s associating buckets to these groups based on tags.

So if next year bucket let’s say 2027 is created with the same tag, Clumio will automatically recognize it and have it fall under the same protection model even before you run the skill. And the other important part, cloud has always stopped and asked us for approval before it applies any of the changes. So since everything looks good, let’s go ahead and approve it.

All right. Now it’s applying the Terraform. And after that, it’s going to call into the MCP to verify that the changes were applied correctly.

So here we see that the verification is complete. It’s writing a report which I’ll touch upon in a moment. And there we go. The apply has completed and we’ve gotten the exact result we were aiming for. And as a bonus, the skill has also created a report that summarizes everything we’ve done as a part of this workflow.

So let’s take a quick look at that report. So here’s our production score. We are now 100% protected and covered by a sensitive data or non sensitive. We’ve gotten a nice summary of what was protected broken down by data sensitivity, as well as a breakdown of the protection by both prefix and bucket. So the same workflow that made the protection changes was expanded to give us a report now, which we can share with stakeholders such as upper management, the application team, etc..

And because cloud is using context from our session, weather report, dashboard and email, whatever it is that you need, it’s all up to you and you can expand the skill accordingly. And as a final note, this is the Terraform configuration that’s generated. And that cloud applied is essentially from local variables. And then from your policy resource at the for each loop that will create three policies.

And there’s also a KML protection Group resource that also has a for each loop that will create three protection groups. So in closing we started by asking where our sensitive data was. This helped us to find it and understand the risk to meet the protection gap and help close it. Then we turned that sequence into a scale. One request that created the Terraform that we wanted and our intended policies, and asked us for approval and verified everything was applied correctly.

We even got a report when it was all said and done.

And if you ever need to restore, recovery is just as easy. None of the underlying capabilities are new on their own with understand cloud risk, cleaning up attacks and recovers data.

What’s new is how easily an agent can bring those capabilities together, and how a skill can make that collaboration repeatable.

All right. And that’s the end of the demo. So, on behalf of Waze and Clumio , thank you again for attending.
