My friends, colleagues, and I use AI coding agents almost daily with OpenCode. The problem is when someone needs help mid-session, you either share a screen or hover over their shoulder. There's no way to just jump in. This process reminded me a bit of 'rubberducking'.
I built Qwack to help solve this problem. One person hosts the agent and others join. Everyone shares the same context, sees the same output, and can send prompts to help steer the agent. The host's machine runs everything, the server is just a WebSocket relay.
I built Qwack to help solve this problem. One person hosts the agent and others join. Everyone shares the same context, sees the same output, and can send prompts to help steer the agent. The host's machine runs everything, the server is just a WebSocket relay.
GitHub: https://github.com/qwack-ai/qwack