Hello, World

Introducing KRAIT — an open-source self-evolving AI agent built on Elixir/OTP with structural security guarantees.

Why We Built KRAIT

The current generation of AI agents can write code, browse the web, and manage files. But they do it blind — with no structural safety guarantees, no audit trail, and no kill switch.

We built KRAIT because we believe self-modifying agents are inevitable, and the safety mechanisms need to be structural, not aspirational. Every mutation KRAIT produces is cryptographically proven, validated by a Rust NIF, and gated by human review.

What Makes KRAIT Different

KRAIT is not another wrapper around an LLM. It is a full-stack agent runtime built on Elixir/OTP with three key properties:

• Self-evolving: The agent writes, tests, and proposes its own skill modules
• Structurally safe: Seven security rules enforced at the AST level by the Narsil Rust NIF
• Fully auditable: Every evolution is a PR with a diff, test results, and security analysis

What's Next

We are releasing KRAIT as open source under the Apache 2.0 license. The initial release includes the core runtime, the Narsil NIF, and the evolution pipeline. We will be publishing detailed architecture documentation, security analysis, and guides for building your own self-evolving agents.

Follow the project on GitHub to stay updated.