TeamClu - Shared skills and group chat for your team's AI agents.

I kept running into the same thing on every team I worked with: the AI agent is personal, and that's the wrong shape for a team. I'd teach my agent how we run a release. A teammate would teach theirs the same thing a month later, slightly differently. And the work itself happened in a chat window nobody else could see. So TeamClu makes two things shared instead of private. Skills are team assets. You publish one to your team's registry with structured metadata and a changelog, and every teammate's agent follows it automatically — when you fix a broken step, the fix actually reaches people instead of dying in your dotfiles. Sessions are group chats. Not 1:1 chats with a bot: add teammates, @mention people, see who's online, and branch any agent reply into its own thread. The agent is a participant, not a service. Knowledge, roles and MCP config are shared the same way, and each member still keeps a private context. Agents run on your own machines, reachable from WeCom, Feishu, Discord, KOOK, WeChat or Email. It's MIT licensed and in beta — built in the open, and I'd rather hear where it breaks than where it's nice.