Jack Dorsey's 'Buzz' throws AI agents into your group chat to kill Slack

Jack Dorsey has a new weapon in the workplace chat wars, and it puts bots on equal footing with your coworkers. On Tuesday, the Block CEO unveiled Buzz, an open-source group chat platform designed to merge human conversation with AI agents in a single stream. The pitch is simple: why toggle between Slack, GitHub, and a terminal when a decentralized app can host your stand-up, your code reviews, and your automated pull request summaries in one window? Dorsey described the tool on X as being “model-agnostic, decentralized, self-sovereign, and open source.”

Built by Block—the parent company of Square, Cash App, and Tidal—Buzz isn’t just a skunkworks side project. It takes direct aim at the fragmentation plaguing teams that rely heavily on AI agents. The interface looks remarkably familiar to anyone who has used Slack, but the native ability to manage GitHub projects and issue commands to AI models without leaving the room is the differentiator here. Because the code is fully available, teams don’t have to wait for a vendor to ship a feature; they can fork the repo and build it themselves.

Dorsey isn’t the only one bored with standard chat ops. Paradigm CTO Georgios Konstantopoulos just showed off a similar concept called Centaur, which he calls a “virtual employee” that lives inside Slack or an API. Konstantopoulos noted that in an enterprise context, self-hosting is critical for security, and users want this capability to surface exactly where they already talk. “There’s a lot of room for improvement for agents that live in Slack and can do more work than just coding for teams,” he wrote.

The catch? Buzz is extremely young. The team openly admits the software is in its “early stages,” meaning it’s likely too raw for a mission-critical migration right now. For a bootstrapped startup that hasn’t yet pledged allegiance to a specific chat ecosystem, it’s a fascinating signal of where workplace tools are heading. The desktop app is available now for macOS, Windows, and Linux, with the source code dumped on GitHub for anyone brave enough to spin up a self-sovereign instance.