Reviewable Agent IDE ORG-II | Generated by AI
Question: What is the GitHub repo yorgai/ORG2?
Answer:
ORG-II is an open-source, Cursor-style agent IDE built on a fundamentally different philosophy: reviewability and control over raw coding speed.
Key differentiators from Cursor/Claude Code:
Architecture
- Built with Rust + Tauri — local-first, <100MB on disk
- TypeScript (54%) + Rust (39%) split: frontend in TS, agent runtime in Rust
- Uses your existing API keys (no new subscription)
The core insight it’s exploring — Cursor/Claude Code treat agents as “outsourced assistants”: useful for output, black-box for audit. ORG-II treats agents as persistent, observable colleagues in a structured org:
- Agent trajectory livestream + replay — every agent session is replayable. You can audit exactly what the agent did, step by step. “AI blame” (like
git blamebut for agent actions) - Cross-session memory + cross-agent knowledge sharing — agents accumulate context across sessions, not stateless per-chat
- Long-running + scheduled sessions — agents can run overnight, auto-restart, react to CPU/RAM availability
- GUI, CLI, terminal, Git, browser, LSP, database tooling all integrated
- Browser Use + Computer Use via native sidecars (
agent-browser,peekabooon macOS)
Status: v1.0.1.21, released June 20, 2026. 737 stars, AGPL-3.0.
Why it matters to you: This is philosophically aligned with what you’d build — it’s agent-architecture-first, not LLM-wrapper-on-editor. The “org-level alignment” framing (agents as colleagues with accountability, not black-box assistants) is the missing layer above what Claude Code/Cursor give you. The replayable execution traces are particularly interesting for debugging agent failures, which is a real pain point at scale.
Worth watching if you’re building CLI agents (ww, zz) and care about auditability of what the agent actually did.
References: