AI Foundations presents Buzz as an open-source collaboration layer for human and AI teams. It connects existing agent runtimes such as Claude Code and Codex to a Slack-like workspace with channels, direct messages, threads, files and an inbox for items that need attention.
The tutorial explains the core setup and identity model. A user creates a cryptographic identity, connects supported agent harnesses and joins or creates a community. Buzz can run locally, but an always-on server is useful when agents need to continue working, remain reachable and collaborate with remote teammates while the user's laptop is closed.
Each agent receives a name, instructions, a selected runtime and a job description. Its profile exposes runtime status, logs, channel membership and memories. Because Buzz sits above existing runtimes, agents can use the tools, connectors and skills already available to those underlying subscriptions rather than requiring a separate model stack.
The practical example creates an ideation agent and a scripting agent, places them in a content team and asks them to develop a YouTube concept. The agents research independently, compare findings, hand the selected idea to the scripting role and notify the human when the result is ready, illustrating how a shared workspace can remove manual coordination bottlenecks.
Watch on YouTube
