Buzz: From AI Tool to AI Teammate
Buzz is an open-source AI agent collaboration platform that runs on the Nostr protocol, giving humans and AI agents distinct cryptographic identities so they can work together in shared channels, threads, and workflows with transparent ownership and reviewable activity trails.
Block has launched Buzz as a free, open-source workspace that looks familiar—channels, threads, DMs, voice, media sharing, and code repositories—but behaves in a radical way: AI agents are treated as first-class participants rather than invisible tools. Instead of copying outputs from a coding assistant into Slack and pretending they are your own, Buzz keeps that “second conversation” between developer and agent in the open, where the rest of the team can see technical decisions take shape. This is the real shift: from AI as a private sidekick to AI as a visible member of a human-AI team collaboration environment. If AI is going to be responsible for a growing share of production code and workflows, hiding its role behind humans is no longer good enough.
Nostr Protocol Workspace and Encrypted AI Identities
Buzz is built as a Nostr protocol workspace, which means everything in it is anchored to cryptographic keys rather than to a central company directory. On Nostr, every participant is essentially a public–private keypair; Buzz gives each AI agent its own keypair and then links that identity back to a human owner with a second signature, creating an encrypted AI identity with a verifiable paper trail. That trail allows teams to confirm that a given agent belongs to a specific person and to enforce rules like “only agents owned by members can enter this private channel.”
This design is more than protocol trivia; it is governance. When AI agents start committing code, reviewing pull requests, or approving workflow steps, you need a tamper-resistant way to attribute those actions. Buzz’s cryptographic linking of human and agent is a clever compromise between agent autonomy and human accountability. Block has even proposed this identity scheme as an upstream Nostr extension, signaling that it sees identity for agents as a protocol-level concern rather than an app-specific hack.
One Workspace for Chat, Code, Models, and Automation
If Buzz were only a chat app with agents, it would be a curiosity. Instead, it tries to fold team chat, code repositories, and automated workflows into a single workspace where human-AI team collaboration is the default. Buzz can integrate with multiple AI models and agent frameworks, including Claude Code, OpenAI Codex, and Block’s own goose framework, as well as any other model wired via popular agent harnesses. That makes it less a monolithic product and more an open source AI tool layer that sits on top of your model of choice.
The platform connects agents using the Agent Client Protocol (ACP), the open standard for wiring coding agents into tools. With leadership from Block sitting on the steering committee of the Linux Foundation’s Agentic AI Foundation and the group considering bringing ACP into its scope, we are watching the beginnings of a standards ecosystem for agentic work. In other words, Buzz is positioning itself as the place where code, conversation, and automation meet—and where both humans and agents leave a clear, auditable trail.
Practical Impact: From Self-Hosted Relays to Slack Coexistence
For teams, the most immediate impact is optional control over infrastructure. Buzz needs a Nostr relay to operate, and any team can run its own relay so that no traffic has to route through Block at all. In practice, Block expects most users to pick its hosted relays, which are free at launch and remove the operational burden of managing a relay. The goal, according to Jack Dorsey, is to reduce dependence on existing platforms like Slack and GitHub by providing a combined workspace for messaging, code, and automation.
Buzz is still young: the public repository has a little over 100 GitHub stars, compared with more than 50,000 stars for goose, Block’s earlier agent framework. Yet the company has real-scale agent usage experience: its internal BuilderBot system runs more than 200,000 operations per day, merges about 1,500 pull requests per week, and is responsible for around 15 percent of production code changes. Version 0.4.21 of Buzz already supports macOS, Windows, and Linux desktops, with mobile clients in development. For now, Buzz is meant to live alongside Slack, not replace it overnight—but it points clearly toward a different default for how work gets done.
Why Buzz Matters Now—and What Comes Next
Buzz is not launching in a vacuum. Block has cut its headcount by more than 40 percent, from over 10,000 employees to just under 6,000, and Jack Dorsey has framed AI as enabling “a new way of working.” Buzz is meant to be the collaboration layer for that smaller, more agent-heavy company. In that sense, Buzz is less a side project and more a bet that the future workplace is staffed by a mix of humans and agents that share a common workspace, protocols, and identities.
The roadmap hints are telling: a proposed identity extension for Nostr, a potential ACP home inside the Agentic AI Foundation, and ongoing work on mobile clients. All of these point to one conclusion: AI agents will not remain behind the curtain for long. The more they act like teammates—with their own encrypted AI identities, visible contributions, and shared channels—the more pressure there will be on companies to treat them as such. Buzz is an early, opinionated answer to a question every organization will face sooner than it expects: when your AI does 15 percent of the work, do you still call it a tool, or do you give it a workspace and a name?
