Key takeaway
- Cloud native applications run as loosely coupled services in any cloud. See the CNCF definition, building blocks, twelve factors and how to start building.
A cloud native application is software built to run in public, private or hybrid cloud environments as loosely coupled services that are deployed, scaled and updated in a repeatable, automated way. It describes how software is designed and operated, not a product you buy, and it usually combines containers, microservices and automated delivery pipelines.
This guide is for architects, platform engineers and technology leaders deciding how to build or modernise applications. It covers the CNCF definition, the building blocks, the twelve-factor practices, a build path and how AI agents change the work. For design patterns, read our guide to cloud native architecture patterns and design, or see our cloud native applications page.
What is a cloud native application?
The CNCF Cloud Native Definition (version 1.1) says cloud native practices "empower organizations to develop, build, and deploy workloads in computing environments (public, private, hybrid cloud) to meet their organizational needs at scale in a programmatic and repeatable manner". It describes the approach as "characterized by loosely coupled systems that interoperate in a manner that is secure, resilient, manageable, sustainable, and observable."
The same definition says cloud native technologies "typically consist of some combination of containers, service meshes, multi-tenancy, microservices, immutable infrastructure, serverless, and declarative APIs", and notes the list is not exhaustive. AWS puts it more simply: "Cloud native is the software approach of building, deploying, and managing modern applications."
Cloud native is not the same as cloud hosted. An application copied unchanged onto a virtual machine is hosted in the cloud. A cloud native one is designed so that its services can fail, scale and ship independently.
What are the building blocks of cloud native applications?
| Building block | What it is | Why teams use it |
|---|---|---|
| Microservices | Small, independent components that together form a complete application (AWS) | Loose coupling lets the application keep running if one microservice fails (AWS) |
| Containers | A package of a microservice and its dependencies; AWS calls it the smallest compute unit in a cloud native application | The service runs consistently across environments |
| Orchestration | Kubernetes is described as a portable, extensible, open source platform for managing containerized workloads and services | Declarative configuration and automation |
| CI/CD | Developers merge changes often with automated checks; delivery keeps services ready to deploy (AWS) | Smaller changes carry less release risk |
| Infrastructure as code | Environments defined in files kept in version control | Environments can be recreated and reviewed like software |
| Serverless | The provider runs the servers and the platform scales the code | Less infrastructure to manage for suitable workloads |
| Monitoring and logging | Telemetry from every service | CNCF lists "observable" as a property of cloud native systems |
What do the twelve-factor practices add?
The Twelve-Factor App, written by Adam Wiggins, is a methodology for building software-as-a-service apps that are scalable, maintainable and portable on modern cloud platforms. It applies to any language and any mix of backing services, and it draws on experience deploying apps on the Heroku platform.
- Codebase
- Dependencies
- Config
- Backing services
- Build, release, run
- Processes
- Port binding
- Concurrency
- Disposability
- Dev/prod parity
- Logs
- Admin processes
Use the list as a review checklist for each service. Practices such as keeping configuration out of code and making processes disposable are what allow a platform to replace and scale instances safely.
How do you build a cloud native application?
- Adopt DevOps ways of working, so developers and operations share ownership of build, release and run.
- Work with architects to split a monolith along business capabilities, one service at a time.
- Review each service against the twelve factors.
- Package each service as a container and run it on an orchestrator.
- Keep infrastructure as code in version control, such as Git.
- Build a CI/CD pipeline with automated tests at every stage, so testing happens early and often instead of after development.
- Add monitoring and logging before the first production release.
- Automate remaining manual steps and keep improving the pipeline.
Network functions have their own patterns, covered in cloud native network functions, and our guide to CNCF cloud native continuous integration and dashboards goes deeper on pipelines.
How does cloud native compare with a traditional application?
| Area | Traditional monolith | Cloud native |
|---|---|---|
| Unit of release | One large release | Many small services released separately |
| Scaling | The whole application together | Each service on its own demand |
| Infrastructure | Often configured by hand per environment | Defined in code and recreated on demand |
| Failure | One fault can stop the whole application | Designed so one failing service does not stop the rest |
When is cloud native not the right fit?
More services mean more moving parts: distributed debugging, service networking and platform skills all cost effort. A small application owned by one team can be a well-structured monolith. Portability also needs checking dependency by dependency, because a managed database or messaging service can still tie an application to one provider even when the code runs in containers.
How do AI agents change cloud native applications?
AI agents can help build and operate cloud native systems: drafting and testing code, reading telemetry, and proposing a rollback or a scaling change. Because agents act through your APIs, versioned, documented APIs and good observability become what the agent relies on, and policy must decide which actions need a human. For teams exploring agents on DevOps, testing and automation work, see ElixirClaw, the AI agents computing platform.
What should you do first?
- Pick one application whose releases are slow or risky.
- Draw its service boundaries and find the first service to extract.
- Containerise that service and review it against the twelve factors.
- Put it behind a CI/CD pipeline with automated tests and monitoring.
- Review the result before extracting the next service.
To map one decision your agents make across your services, and the data and policy it depends on, book a Context Workshop. See what happens in the workshop.
Frequently asked questions
What is the difference between cloud native and cloud hosted?
Cloud hosted means an application runs on cloud infrastructure. Cloud native means it is designed as loosely coupled services that are deployed, scaled and updated in an automated, repeatable way.
How does the CNCF define cloud native?
The CNCF says cloud native practices let organisations develop, build and deploy workloads in public, private and hybrid cloud environments at scale, in a programmatic and repeatable manner, using loosely coupled systems that are secure, resilient, manageable, sustainable and observable.
Do cloud native applications need Kubernetes?
Not by definition. The CNCF lists containers, service meshes, microservices, immutable infrastructure, serverless and declarative APIs as typical technologies, "some combination" of them. Kubernetes is a common platform for managing containerised workloads.
What are the twelve factors?
They are codebase, dependencies, config, backing services, build, release and run, processes, port binding, concurrency, disposability, dev/prod parity, logs and admin processes.
Is a monolith ever the right choice?
Yes. A small application owned by one team may not need the extra moving parts of many services, so weigh the operating cost before splitting it.
Follow XenonStack on LinkedIn: XenonStack on LinkedIn for new guides on agentic AI, data and platform engineering.
Ready to put AI agents to work? Explore ElixirClaw, the agentic OS platform to build, orchestrate and govern AI agents.
Explore the XenonStack platform family: ElixirData · ElixirClaw · ElixirHub · NexaStack · MetaSecure · Neural AI · AgentGRC · AgentSRE · AgentAnalyst · AgentInstruct · AgentRAI · AgentScrum · AgentSearch · AgentForce · AI Aviator · SAIF Aviator · CohortIQ · XenonStack AI · XenonStack UAE