james_library

This is a Rust/Python research stack with internet access, focused on reproducible experiments through voice conversations. We wanted to build something that would give everyone a free, local, and customizable 20th-century Bell Labs on their computer. The platform combines ZeroClaw (Rust runtime) for orchestration and tool policies, James Library (original research articles/prototypes) for sessions and experiments, and a security-first agent model that treats external text as untrusted. The goal is simple: go from idea to experiment to artifact, without black boxes, and more customizable than most cloud-based multi-agent platforms. For reference, we’ve been using this daily for a few years as an R&D branch for Vers3Dynamics, and recently decided to open-source it.

I’ve been working on a multi-agent academic research lab, and I wanted to share it here today primarily to give a massive shoutout to the developers behind ZeroClaw. When designing the architecture for this, I needed an autonomous agent runtime that was lightweight, entirely agnostic, and highly secure for local execution. ZeroClaw’s pure Rust implementation provided exactly the zero-overhead foundation the project required. Because they solved the core runtime execution so elegantly, I was able to spend my time building the higher-level orchestration on top of it—like the retrieval graph engines, the Godot-based client interface, and the physical hardware/robotics bridges. If any of the ZeroClaw maintainers are reading this: thank you. Your repository is a cool example of open-sourcery, and the R.A.I.N. Lab literally would not exist in