emaclaude

An agentic coding system that runs directly within Emacs. Three Claude code sessions coordinate across split buffers — one plans, one codes, one reviews — connected through a lightweight Rust CLI and an Elisp orchestrator.

The review agent scrutinizes the implementation and feeds corrections back to the coder. The loop continues until the reviewer approves twice consecutively — a safeguard against hallucinated approvals. Once cleared, a Magit diff view appears for annotation, and you can submit feedback directly to the coding agent for refinement before creating a PR from within Emacs.

No Electron app needed.