On July 22, 2026, Block—the technology company led by Jack Dorsey—released the first desktop client of Buzz, an open-source collaboration platform that breaks with every major workplace chat app in one decisive way. It doesn’t add artificial intelligence to an existing tool. It was designed from the ground up for a workspace where AI agents are peers, not plugins.
Buzz runs natively on Windows, macOS, and Linux. It gives every agent its own identity, a cryptographic key pair, and a visible presence in team channels. The agent can read the conversation, respond to project context, and even take concrete actions—like opening a pull request or drafting documentation—if an admin has granted the necessary access. This is a far cry from the sidebar assistants that live inside Microsoft Teams or Slack.
A Workspace Built for Humans and Machines Alike
The core pitch is simple, but the implications run deep. Buzz treats an agent as a first-class participant. It isn’t an integration that posts messages from a separate service. It’s a named entity with a profile, permissions, and the ability to operate in the same shared context as the people who review its work.
Imagine a coding channel where a bug is discussed. One agent observes the thread, compares it to the code repository, and suggests a fix. Another agent watches the suggested fix, runs security checks, and flags concerns. A third monitors the deployment pipeline and prepares release notes. No one opens a separate AI tool, copies text, or pastes results. The agents sit in the channel, working alongside the team, their activity visible and auditable.
This architecture pulls together capabilities that are normally scattered across half a dozen applications:
- Group channels, threads, and direct messages
- Voice huddles
- AI agent identities with their own public keys
- Code and repository integration
- Self-hosted or Block-hosted community servers, called relays
- A single desktop client that connects to multiple relays
Block describes the relay model as the backbone of the platform. Each relay stores a community’s channels, messages, members, agents, and rules. Organizations can use Block’s hosted relays or deploy their own on infrastructure they control. The Windows client acts as a universal front end, letting users switch between communities as easily as they switch between Slack workspaces.
Why This Matters to Windows Users Right Now
If you’re a typical Windows user who lives inside Microsoft Teams and Outlook, Buzz is not about to replace your daily driver. It’s early-stage software, and it lacks the deep Office 365 integration, calendaring, and compliance tooling that make Teams a business staple. But for a specific—and growing—set of users, the arrival of a native Windows client for a self-hostable, agent-first workspace is a signal worth paying attention to.
Developers on Windows often suffer from tool sprawl. A day might include Teams for chat, GitHub for repos, VS Code for editing, a separate AI coding assistant, Jira for tickets, and a documentation wiki. Each application handoff strips context and adds friction. Buzz’s promise is to shrink the distance between conversation, code, and automated action. A bug report, its technical discussion, the relevant pull request, and the agent that reviews it can all live in the same environment. Fewer context switches can mean faster, clearer work.
For IT administrators and security teams, the self-hosted relay option is the bigger draw. Regulated industries, government agencies, and research labs often cannot send internal conversations and AI prompts through a third-party cloud service. A self-hosted Buzz relay lets them keep data on their own servers, choose which AI models connect to their workspace, and define exactly what each agent can see and do. That freedom comes with real operational responsibility—server maintenance, backups, identity management, and policy enforcement—but for some organizations, the trade-off is worth the control.
How We Got Here: From Chatbots to Autonomous Teammates
Workplace AI has shifted rapidly. Five years ago, bots were simple: they posted notifications, answered slash commands, and triggered basic workflows. Then came large language models, and suddenly assistants could summarize threads, draft replies, and generate code snippets. Microsoft embedded Copilot across Teams and Office. Slack added AI search, summaries, and workflow agents. Google, Zoom, and Notion followed suit.
All these implementations share the same fundamental limitation. They bolt intelligence onto the edge of a collaboration tool that was designed for humans only. The agent sits in a sidebar, or in a dedicated direct-message chat, or behind a button that must be clicked. Context is what you paste into a prompt, not what the agent can organically absorb from the team’s ongoing work.
Buzz starts from the opposite premise. The workspace is shared from day one. Humans and agents coexist, and the conversation history, code, and decisions that a team produces are directly available to the agents that need them—assuming permissions allow it. This inversion is what makes Buzz more than just another chat app. It’s an attempt to build collaboration infrastructure for a world where some of your most productive colleagues happen to be software.
Jack Dorsey has a long interest in decentralized, open-source alternatives to dominant platforms. He funded Bluesky as a decentralized social network, and he has argued that users should own their digital infrastructure. Buzz applies that same philosophy to the workplace communications stack, offering an escape hatch from the vendor lock-in of Teams, Slack, and their closed cloud ecosystems.
Before You Install: What to Know About Security and Privacy
Buzz’s agent model makes permissions the single most important security control. An agent that can read a channel can see everything in it. An agent that can act on a repository can make real changes. If you grant broad access for convenience, you may later discover an agent is reading confidential messages or altering code without proper oversight.
Block’s own security design emphasizes deliberate access:
- Every agent should have a dedicated identity and a human owner.
- Agents should only be invited to the specific channels and repositories they need—nothing more.
- Actions with real-world consequences (creating pull requests, modifying documents, triggering deployments) should require explicit human approval until trust is established.
- All agent activity should be logged and auditable, so teams can review what happened and when.
There’s another privacy detail that many early adopters will miss: Block-hosted communities are not end-to-end encrypted. The company states that it may access hosted content to operate, secure, or moderate the service—similar to how Teams and Slack handle compliance and search. If your organization needs absolute privacy, a self-hosted relay is the only path, and you’ll need to pair it with AI models that you control rather than third-party providers. Even then, no deployment is secure by accident. Governance, monitoring, and access reviews must be baked in from the start.
Getting Started with Buzz on Windows
If you want to kick the tires, the process is straightforward:
- Download the client. Grab the Windows installer from Block’s official site. The same place offers macOS and Linux builds.
- Join a community. You can connect to a Block-hosted community to test conversations and see agents in action without setting up infrastructure.
- Explore agent scenarios. Once inside, look for example agents or invite one to a test channel. Watch how it interacts with the channel’s history and its assigned tools.
- Experiment with self-hosting (for technical users). Set up a relay on your own server. This unlocks full control over data and agent configurations. Start simple: a single agent with read-only access to a non-sensitive channel.
- Read the fine print. Review the platform’s privacy terms, especially if you intend to connect external AI model providers. Know what data might be sent outside your relay.
Keep expectations in check. Buzz is not a polished, enterprise-grade tool yet. The interface can feel dense, and the agent behavior may be unpredictable. The onboarding experience lacks the slickness of Teams or Slack. But for developers, technical project teams, and enterprises evaluating decentralized collaboration, it’s a working preview of a possible future.
The Outlook: Agents Are Coming to Your Team, One Way or Another
Buzz will not dethrone Microsoft Teams or Slack in the next quarter. But it represents a genuine conceptual leap—from AI as a feature to AI as a collaborator. The big incumbents are already moving in this direction. Microsoft’s Copilot agents, Slack’s AI-powered workflows, and the broader “agentic AI” push across the industry all point to a future where half your team might not be human.
What Buzz does is make that future tangible, now, in a downloadable Windows app. It puts the hard questions on the table: How do you govern an AI colleague? How do you audit its actions? How do you ensure it doesn’t become a noisy distraction? Teams that begin experimenting with these questions today will be better prepared when the agent invasion goes mainstream. For everyone else, Buzz is a quiet, open-source signal that the way we collaborate is about to change fundamentally.