Meldwork
A local-first desktop app for macOS that coordinates several installed AI coding-agent CLIs -- Codex, Claude Code, Gemini CLI, and others -- from one reviewable workspace, capturing each agent's findings as structured evidence behind a human approval gate.
Meldwork connects AI coding-agent CLIs already installed on a machine, including Codex, Claude Code, Gemini CLI, Qwen Code, Kimi Code, and several others, into a single workspace instead of separate terminals. A task snapshot is frozen and sent to every selected agent, and Meldwork can run a direct session with one agent, concurrent responses where several agents answer the same task independently, or an auto-discussion mode where agents propose, challenge, and negotiate responsibilities across rounds.
Unlike a terminal multiplexer that just shows raw output side by side, it records each agent's findings as structured evidence -- finding, evidence, decision, and disposition -- and keeps a human adoption gate before any file in the workspace changes, so multi-agent sessions stay inspectable and reusable later.
It is a macOS desktop app for Apple silicon, distributed as a direct download.