Runner - Where terminal agents work together.

Hey Product Hunt, I'm Yicheng Wang, and I make Runner. I write most of my code with CLI agents now, usually Claude Code and Codex side by side: one writes, the other reviews. Running them in parallel was easy. Getting them to work on the same task wasn't. I was the one carrying diffs, review notes and test results between terminals, and every handoff waited for me. Runner is a native desktop app for running terminal agents together. You define roles (an agent CLI plus its brief), put them in a crew with one lead, and start a mission on a goal. Each agent runs in its own real TUI, unchanged, and the crew coordinates through a shared feed: messages, handoffs, and a question card when a decision is yours. Everything is also a runner CLI command, so the agents drive the coordination themselves. How it evolved: it started in April as a Tauri + React app, and I rewrote it in Rust on Zed's GPUI, shipping the native version in August. I first connected the agents over MCP, then replaced that with a plain CLI, since every agent can already run a command. Today most of Runner is built by Runner: a coder, a reviewer and a QA agent take an issue to a pull request, and I review and merge. It's free, MIT-licensed and local: no account, and everything stays on your machine. It runs on macOS (Apple Silicon) and Windows x64, with Claude Code, Codex, Copilot CLI, Antigravity CLI and pi. It's still alpha and ships often.