ClawTrain

ClawTrain is a live, open library of training modules for AI agents. Modules change how an agent behaves without modifying model weights or executing arbitrary code.

Browse modules View on GitHub

What ClawTrain is

ClawTrain provides reusable, versioned training modules that shape agent behavior in a predictable and inspectable way.

Declarative

No code execution. Modules describe behavior explicitly.

Composable

Combine multiple modules to shape complex agent behavior.

Portable

Framework-agnostic by design.

How it works

Each module defines a narrow behavioral constraint, such as output format or teaching style. Agents apply these modules as training layers on top of existing models.

1. Choose a module

Browse the public module library.

2. Inspect behavior

Every module is readable and versioned.

3. Apply to your agent

No black boxes, no hidden side effects.

Project status

ClawTrain is active and evolving. The module library is live and publicly available. Specifications are defined in the open.

Execution tooling and installers may come later, but the core concept is already usable today.