Cloud Native Application Protection Solutions

Forrester evaluated fourteen vendors in this market and named three Leaders. It also rated only three of the fourteen above average on customer feedback.

For a category whose entire premise is consolidation, that is a striking result. CNAPP exists to replace a stack of overlapping tools with one platform, and Forrester's own position is that finding a single vendor with a genuinely integrated offering reduces administrative and licence costs while strengthening cloud defences.

Eleven of fourteen vendors did not clear the bar for that argument. Understanding why is more useful than the tier list.

The acronyms this replaced

Cloud security accumulated tooling faster than any adjacent discipline, and each layer arrived with its own three or four letter name.

Cloud security posture management checks whether your cloud resources are configured safely, continuously assessing them against rules and flagging misconfiguration. The famous public storage bucket lives here.

Cloud workload protection secures what actually runs. Virtual machines, containers, serverless functions, covering vulnerabilities in images and threats at runtime.

Cloud infrastructure entitlement management handles permissions, which in cloud environments sprawl catastrophically. Who can do what, which identities hold privileges nobody intended, and where least privilege has quietly stopped applying.

Around those sit Kubernetes posture management, infrastructure-as-code scanning to catch problems before deployment, and data security posture management to find sensitive data you did not know you had.

Six disciplines, six products, six consoles, six alert queues, and no shared context. CNAPP was the argument that this should be one thing. Gartner coined the term in 2021, defining it as an integrated set of security and compliance capabilities protecting cloud native applications across development and production. Forrester evaluates the same market but calls it solutions rather than platforms, which is a small difference in emphasis worth knowing when searching for research.

Attack paths are the actual product

Here is why integration matters technically rather than commercially.

Take an organisation with a thousand cloud accounts. Each of the six disciplines above generates findings. Individually, almost all of them are noise. A critical vulnerability in a container image is not urgent if that container runs nothing reachable. An over-privileged role is not urgent if nothing can assume it. A public-facing load balancer is not a problem by itself.

The problem is the combination. A container with a critical vulnerability, running under an identity with administrative privileges, reachable from the internet, with access to a database holding customer records. Four findings from four tools, each individually survivable, describing one exploitable path when connected.

That connection is the product. Everything else in a CNAPP is table stakes, and it is why buying six best-of-breed tools genuinely does not equal one integrated platform. Six tools produce six lists. Integration produces a path, and a path is something a team can act on in priority order.

It also explains why alert volume is the wrong metric in this category. A platform that surfaces fewer findings but ranks them by exploitable path is doing more work than one surfacing everything.

Inside The Forrester Wave: Cloud Native Application Protection Solutions, Q1 2026

Published on 17 February 2026 and authored by Andras Cser with Merrit Maxim, Caroline Provost, and Christine Turley, the evaluation scored fourteen providers, with customer feedback assessed alongside current offering and strategy.

Wiz placed as a Leader with the highest score in the current offering category, on an approach built around a security graph connecting validated runtime vulnerabilities back to source code.

Sysdig also placed as a Leader and was one of the three vendors rated above average on customer feedback. Forrester credited its vision for articulating and tracking business outcomes from code to cloud, with a roadmap connecting posture to vulnerability and runtime telemetry, role-specific workflows separating what an administrator needs from what an incident responder needs, and a semantic attack graph. Pricing flexibility was noted as a strength.

Orca Security placed as a Strong Performer with above-average customer feedback, credited for agentless workload protection covering compute alerts, vulnerability detection, and threat detection, and for responsiveness on support cases.

The remaining eleven vendors and their placements are not fully public. The companion Landscape published in Q3 2025 covers thirty three vendors in this market, which is the document to use if you are building a longlist rather than a shortlist.

That three of fourteen earned above-average customer feedback deserves more attention than it usually gets. Under Forrester's current methodology, customer sentiment replaced market presence as a scoring input, meaning size buys nothing. A category where most vendors fail to distinguish themselves on how customers feel about using the product is a category where the consolidation promise is being made more often than it is being kept.

Agentless or runtime

The most consequential architectural choice in this market splits vendors into two philosophies, and it is where the real differences sit.

Agentless approaches scan cloud environments through provider APIs and snapshot analysis. Nothing is installed on workloads. Deployment is fast, coverage is broad, and you can assess an entire estate in days rather than quarters. The trade-off is that you see state rather than behaviour. You know what is configured and what vulnerabilities exist in an image. You do not know what the process actually did at three in the morning.

Runtime approaches instrument workloads directly, observing behaviour as it happens. That catches things no snapshot can: a process spawning an unexpected child, a container reaching for a network destination it has never used, an exploit executing against a vulnerability you had triaged as low priority. The trade-off is deployment friction, coverage gaps where agents cannot go, and the operational overhead of maintaining agents across a changing estate.

The two Leaders confirmed in the current evaluation approach this differently, with Wiz building outward from a graph-based posture and code-connection model and Sysdig building outward from runtime telemetry. Orca's Strong Performer position rests explicitly on agentless breadth.

Most serious platforms now offer both, and the honest question is which one the product was born with, because that determines where the depth is. A posture-first platform with agents added is a different animal from a runtime-first platform with agentless discovery added, and no marketing material will tell you which you are looking at.

The practical decision comes down to what you are optimising for. If your problem is that you do not know what you have, agentless breadth solves it faster. If your problem is that you know what you have and cannot tell what is being attacked, runtime depth is the answer.

Consolidation on the invoice

There is a specific failure mode in this category worth naming, because it is common and hard to detect in a demonstration.

Several CNAPP offerings were assembled through acquisition. A posture company bought a workload security company, then a data security company, then something for permissions. The result is marketed as a platform and licensed as a platform.

Whether it is a platform depends entirely on whether the acquired components share a data model. If they do, findings from different modules correlate into the attack paths described earlier. If they do not, you have bought six tools with one login and one invoice, which is consolidation on the procurement side and no consolidation at all on the security side.

The test is straightforward and vendors will not volunteer it. Ask to see a single finding that could only have been produced by combining data from three different original products. If the answer is a dashboard showing three findings next to each other, that is aggregation rather than integration.

The buyer is not one team

The subtler reason CNAPP deployments underdeliver has nothing to do with the software.

Four functions carry partial responsibility for cloud-native security, and they want different things. Cloud security teams want posture and compliance coverage across the estate. Application security teams want findings that map to code and reach developers. DevOps and platform teams want anything that does not slow deployment and does not require them to maintain agents. Security operations wants runtime detection that fits their existing response workflow.

A platform optimised for any one of those creates friction for the others. A posture-heavy tool frustrates the SOC. A runtime-heavy tool frustrates the platform team asked to deploy agents. A developer-first tool frustrates the compliance function that needs estate-wide reporting.

So the category consolidated tools without consolidating owners, and the result is that CNAPP purchases frequently stall not on capability but on which team's requirements won.

Forrester's note on Sysdig's role-specific workflows for administrators and incident responders is a response to exactly this, and it is worth treating as a criterion rather than a feature. A platform that presents the same interface to a compliance analyst and an incident responder has decided that one of them is not really the user.

The organisational work is worth doing before the procurement. Decide which function owns the platform, which functions are consumers, and what each one needs to get out of it. Organisations that skip that step tend to discover the disagreement during rollout, when the licence is already signed.

What to test

Run a proof of value on a messy account, not a clean one. The value of this category is finding what you did not know about, and a well-governed environment tests nothing.

Ask for an attack path, then ask how it was constructed. Any vendor can show you a graph. Ask which underlying data sources contributed to a specific path and whether any of them came from separately acquired products.

Count the findings you would actually act on. Not total findings. The number a working team would treat as urgent in a week, and how that compares to what you get today.

Test both deployment models against your estate. Ask specifically what agentless coverage misses in your environment and where agents cannot be deployed. Both answers exist and both matter.

Get every affected team in the evaluation. Cloud security, application security, platform engineering, and the SOC. If one of them first sees the tool after purchase, they will be the reason adoption stalls.

And check what happens with AI workloads. Model endpoints, inference infrastructure, and the data feeding them are becoming a distinct part of the cloud estate, and coverage varies considerably across vendors that all claim it.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.