eastreams/loong

What it solves

Loong provides a secure and high-performance foundation for building vertical AI agents. It addresses the need for agents that can operate reliably in real-world environments over the long term, focusing on security, performance, and the ability to evolve without compromising safety.

How it works

Loong is built using Rust to ensure predictable runtime overhead and fast execution. It employs an architecture-governed safety model where outbound effects are modeled as Actions, which must be authorized by a policy before they are executed. To allow for growth, it supports two extension methods: deep, performance-sensitive integrations compiled in Rust, and runtime scripts for faster iteration.

Who it’s for

Developers who want to build specialized AI agents with a strong emphasis on security, performance, and long-term maintainability.

Highlights

• Architecture-governed safety: Uses a policy-based authorization system for all outbound actions.
• Rust-based performance: Leverages Rust for a fast and predictable core execution path.
• Extensibility: Supports both compiled Rust integrations and runtime script extensions.
• Actor-based runtime: Uses an Actor runtime and lifetime model to simplify the management of components.