Skip to main content

Getting Started

Welcome to the Gantral project.

Gantral is an open-source AI Execution Control Plane focused on execution semantics, human oversight, and auditability for AI-enabled workflows.

This guide explains how to explore the project and contribute responsibly.


Who This Project Is For

Gantral is primarily intended for:

  • Platform engineers
  • Infrastructure architects
  • Security and governance engineers
  • Contributors interested in execution control, not agent intelligence

If you are looking to build AI agents, optimize prompts, or experiment with autonomous workflows, Gantral may not be the right project.


How to Explore the Project

Start with the following documents:

  1. Positioning

    • What Gantral is
    • What Gantral is not
    • Category definition
    • Expansion narrative
  2. Architecture

    • Technical Reference Document (TRD)
    • Execution plane model
    • Architectural invariants
  3. Governance

    • Open source philosophy
    • License commitment
    • RFC process

These documents describe architectural intent and design principles.
They are the authoritative reference for evaluating changes.


Repository Overview

The Gantral repository is organized around a specs-first model.

Specifications and design precede implementation.

Code that diverges from documented semantics should be treated as incorrect, even if it appears to function.


Before You Contribute

Before proposing changes, contributors are expected to:

  • Read the architectural invariants
  • Understand the execution model and HITL semantics
  • Review existing RFCs (if any)
  • Confirm that the proposed change aligns with Gantral’s scope

This project prioritizes correctness and clarity over speed.


Participation in the project does not create any employment, partnership, or agency relationship.

Contributions are subject to the project’s license and contribution requirements.