Master Buzz AI in 28 minutes

Buzz: An AI-Native Collaboration Workspace for Agent Orchestration

Overview

In this video, the presenter—a solo developer and AI enthusiast—demonstrates why they consolidated all their AI agents (Claude Code, Codex, Hermes) into Buzz, a new AI-native collaboration platform by Block (Jack Dorsey’s company). The video serves as both a conceptual overview and a practical guide to unifying human and AI workflows.

What Is Buzz?

• Built on Nostr, a decentralized messaging protocol based on cryptographic key pairs.
• Messages are signed locally, sent to relays, and forwarded to communities and agents.
• ACP (Agent Communication Protocol) acts as a translation layer, enabling agents to read and act on human commands.
• Users create or join communities (private workspaces) with shared settings and membership.

Key Features

• Channels: Slack-like chat rooms for humans and agents, with support for private, public, and temporary rooms.
• Direct Messages: One-on-one conversations with specific agents for focused tasks.
• Forums: Reddit-style threads for longer-lived discussions with a persistent record.
• Pulse: A social-media-style feed where agents post updates, metrics, or summaries.
• Projects: Centralized view of GitHub repositories, issues, and code changes.
• Workflows: Event-driven automations (e.g., reactions, webhooks, schedules), still experimental.
• Audio Huddles: Voice meetings with transcription, allowing agents and humans to interact and execute action items.

Creating and Connecting Agents

• Users create agents by choosing a harness (e.g., Claude Code, Codex, Hermes, or Buzz’s default) and linking API keys and models (e.g., Opus 5, Fable 5).
• Agents can run locally or on a 24/7 server for persistence, with options for parallelism and sub-agents.
• The presenter demonstrates cross-agent collaboration: Claude Code cross-checking code with Codex, coordinating multiple Hermes agents on different machines, and auto-creating workflows.
• Channels can include both human colleagues and agents, allowing team-wide orchestration.

Tips and Limitations

• Multi-machine setup: Agents are tied to the machine where they were created; a dedicated always-on server is recommended.
• Mobile pairing: Use the QR code to link the app, ensuring you pair with the correct identity to avoid conflicts.
• Costs: API usage can jump 2–3× due to agent collaboration, with bills up to $200/day.
• Bugs: Workflows are unreliable, and relay disconnections occasionally occur.
• Best for: Solopreneurs and small teams; not yet enterprise-ready.

Final Verdict

Despite the rough edges, the presenter moved to Buzz because it solves the recurring context problem. Instead of manually copying information between tools, agents now share project history and hand off tasks autonomously. The pivotal moment came when Claude Code automatically retrieved missing files and credentials from Hermes without human intervention—a task that previously took 30 minutes of manual work.

“Buzz is not for everyone, but it may be the clearest preview of what an AI-first company looks like.”

Key Takeaway

Buzz transforms independent AI agents from isolated tools into a cohesive, context-aware team. While still early-stage, it demonstrates that a connection and context-driven workspace can eliminate copy-paste overhead and unlock a new level of human–AI collaboration. For those ready to embrace the frontier of agentic workflows, Buzz is a compelling step forward.