Get started

Run Kionga on your own machine, then use it end to end: create a project, train and serve a model, and chat with an agent. The local stack is the fastest way to learn Kionga before you deploy it for a team.

Step	Page	What you get
1	Install Kionga	The full platform running locally, with the console at http://localhost:8080
2	Quickstart	A project, a trained and deployed model, a live prediction, and an agent conversation
3	Service endpoints	Connections to MLflow, Prefect, PostgreSQL, Kafka and the other bundled services

Before you begin

You need a computer with Docker Engine and Docker Compose v2, at least 8 GB of memory available to Docker, and 25 to 35 GB of free disk space for the first build. The installation guide lists every requirement.

Local stack versus a team deployment

The local stack uses development credentials and binds every port to your own machine. Those defaults are for learning and development only. To run Kionga for a team, choose a deployment from Deploy and operate and complete the production readiness checklist.

If installation or startup fails, see Troubleshooting.