Which Tools Does the MLOps Guide Use and What Are They For?

The MLOps Guide is built around six named tools: DVC, CML, GitHub Actions, IBM Watson Machine Learning, IBM Watson OpenScale, Terraform, and Cookiecutter. Each one maps to a specific stage of the model lifecycle — versioning, CI/CD, deployment, monitoring, and environment/project setup — and the guide treats them as modular, meaning a team can swap any tool for another that fits its own preferences and use case.

The tools and their roles

The guide's own tool table lists IBM Watson ML as the deployment tool, with the developer and license columns present in the source (the excerpt cuts off before the full license text, so check the site for current licensing terms).

What each tool is actually for

DVC — data versioning and pipelines

DVC handles two jobs in the guide: versioning datasets so data changes are tracked the same way code is, and defining pipelines that connect processing steps. This is the answer to the "data and model versioning" feature the guide lists as a core MLOps property.

CML and GitHub Actions — CI/CD for machine learning

CML and GitHub Actions work together. GitHub Actions runs the automation; CML supplies the ML-specific continuous integration and delivery behavior. The guide splits this into three concrete pieces:

• Implementing tests
• Continuous integration with CML and GitHub Actions
• Continuous delivery with CML, GitHub Actions, and Watson ML

IBM Watson Machine Learning — deployment

Watson Machine Learning is where a trained model becomes a deployed API. In the guide's flow, this is the delivery target that the CI/CD stage pushes to.

IBM Watson OpenScale — monitoring

OpenScale covers continuous monitoring, the last of the five MLOps features the guide defines. It sits after deployment, watching the model in production.

Terraform — environment setup

Terraform provisions the IBM environment. The guide's environment section covers IAM access, setting up the IBM environment with Terraform, and managing the deployment space — so Terraform is the infrastructure-as-code layer beneath everything else.

Cookiecutter — project structure

Cookiecutter generates the starting project layout. The guide's implementation section begins with "Starting a New Project with Cookiecutter," which means it's the entry point before any versioning, CI/CD, or deployment work happens.

How the tools fit together

The guide presents these as a single end-to-end flow rather than isolated tutorials:

1. Cookiecutter scaffolds the project.
2. Terraform provisions the IBM environment and deployment space.
3. DVC versions data and defines pipelines.
4. CML + GitHub Actions run tests, integration, and delivery.
5. Watson Machine Learning serves the model as an API.
6. OpenScale monitors it in production.

The guide states the architecture is modular, so a team can replace any tool at any point based on project preferences and use cases. That matters if you already have, say, a monitoring stack — you can keep it and follow only the parts of the guide that apply.

Where to start

If you want to use the guide, the order above is also the reading order it implies: concepts first (Principles of MLOps: versioning, automation, feature storing, CI/CD, continuous monitoring), then the implementation guide, then environment, versioning, CI/CD, and monitoring. The guide also references a full example end-to-end project built specifically for it, which is the fastest way to see the tools working together before adapting them to your own setup.