openrigor

OpenRigor is a clean-room, open-source implementation of a persistent AI-agent workspace. It is a chat-first product that can plan work, run tools in an isolated workspace, stream progress, survive disconnects, and return inspectable files. The repository is an implementation workspace created from the ClawsGO public-product research snapshot, and it does not copy ClawsGO code, branding, screenshots, text, or protected visual assets. The goal is behavioral and architectural equivalence where the behavior is publicly observable.

The current status includes a completed research baseline and a working local alpha with real Codex app-server execution, streamed answers, reasoning, tools, and files, SSE replay, stop/interrupt, and same-thread follow-up turns. The interface is an original desktop-first operator console with preview-before-send examples, Light/Dark/System themes, Codex-style Sol/Terra/Luna and low through max controls, a separate Ultra multi-agent mode, knowledge and skill selectors, structured reasoning and tool output, and a browser-side recent-task list. Scope is one local user and one configured workspace. Recent conversations can restore their complete in-memory run chain, but runs and events are still process-memory only; attachments, approvals, queued follow-ups, and restart-safe server history are not implemented yet.