The two products

graal (GRAAL.SYSTEMS SAS) publishes two products, both open source, under the Apache 2.0 license: graal Data Platform and graal Agentic Platform. The two platforms are independent and complementary. This documentation is the Data Platform’s; the Agentic Platform is presented on its own page.

graal Data Platform

From source to model in production, on your infrastructure.

Who it is for: data teams (data engineers, data scientists, analysts), CIOs and platform teams.

What it brings together:

connectors to your databases, files and APIs;
low-code pipelines, exported as code;
orchestration and compute: jobs, workflows, scheduling, distributed Spark, GPUs (Jobs and workflows);
an Apache Iceberg lakehouse and SQL with Trino (Data and SQL);
notebooks: Jupyter, VS Code and RStudio (Workspaces);
ML and generative AI: experiments, registry, serving, LLM gateway, RAG (Machine learning);
governance: rights, audit, secrets, costs (Security and governance);
agent-driven operation: an MCP server and one service account per project (AI agents (MCP)).

Where to find it: the graal Data Platform product page presents the overview and each module; the Run by your agents page presents its MCP server. In this documentation, the capabilities overview gives the detail, family by family.

graal Agentic Platform

Your AI agents move your processes forward, under measured guarantees.

Who it is for: CIOs, IT teams, software and business teams that hand work over to agents, and integrators.

What it carries:

workflows of agents and people: software delivery, IT operations, joiners and leavers, data changes, with controlled releases to production (release train);
any agent (Claude Code, Codex, Gemini CLI…) behind a single protocol;
guarantees that refuse: tests actually run, changes kept within what is authorised, no secret, measured facts; and a re-authenticated human approval before any write to a third-party system;
a disposable sandbox, a capped model key and a budget per run, a model gateway and a cost ledger, a tool catalog (MCP) with no tool credential handed to the agent;
measured evidence and an audit trail.

Where to find it: the graal Agentic Platform product page details each point: workflows, any agent, guarantees, models and budgets, evidence. For a demonstration or a technical conversation, write to us.

Independent

Each one can be installed, used and bought on its own: the Data Platform is driven by any compatible MCP client; the Agentic Platform also carries processes that involve no data (joiners and leavers, operations, software delivery).

Complementary

The two platforms complement each other in three ways:

The Agentic Platform’s agents operate the Data Platform through its MCP server, under a service account limited to one project; every call is traced on both sides.
A data change proposed by an agent is governed end to end: branch, guarantees (tests run, data quality on a sample, measured facts), human approval, application through the Data Platform’s API, evidence attached to the ticket.
A common foundation: the company’s SSO (OIDC, SAML, LDAP), the same audit conventions, the same installation (Helm, no cluster-administration right, restricted pods), on your infrastructure.

On the Data Platform side, point 1 is detailed on the AI agents (MCP) page: the server is the same for every client, and the service account only has roles on the projects assigned to it. Point 3 is detailed on the Installation and Security and governance pages.

License

Both platforms are open source, under the Apache 2.0 license. The text of the license is published by the Apache Software Foundation: Apache License, Version 2.0. What the publisher offers on top of the code (publisher subscription, pilot, graal Services) is presented on the Pricing page.