Skip to content
graal — documentation

Overview

graal is a self-hostable data & AI platform: it installs on your Kubernetes and takes you from exploration to production, without your data leaving your infrastructure. Your teams drive it from the console, and your AI agents through the open MCP protocol, with the same rights.

Where to start

Section titled “Where to start”
You want to… Go to
understand the vocabulary before the rest Concepts and vocabulary
a first job that runs, today Quickstart
see every capability at a glance Capabilities overview
install the platform on your cluster Installation
connect an AI agent AI agents (MCP)
answer your CISO Security and governance
understand why it does not start Troubleshooting

What graal does

Section titled “What graal does”
  • Connectors: relational databases, files, S3, SFTP, REST APIs, Hadoop and SAS, with credentials kept as project secrets.
  • Low-code pipelines: draw a pipeline, export it as Pandas or PySpark code, turn it into a job.
  • Jobs and workflows: Python, Spark, SQL, bash, notebooks; graph workflows, cron, event triggers; live logs.
  • Lakehouse and SQL: Apache Iceberg tables on your S3, SQL with Trino, a catalog of layers, databases, tables and fields.
  • Notebooks: Jupyter and VS Code in the browser, behind your SSO.
  • Machine learning: experiments and model registry, compatible with the MLflow client.
  • AI agents (MCP): read and write tools under the agent’s own account, no deletion tool enabled by default, every action attributed and notified.
  • Governance: roles and rights per project, groups, audit, costs per project.

The detail, family by family, with the pages that cover each one: capabilities overview.

Go further

Section titled “Go further”