Block, the company led by Jack Dorsey, released Buzz on July 21, 2026—a free, open-source collaboration app for Windows, macOS, and Linux that puts AI agents in the same channels as human teammates. The desktop-first platform treats software agents as persistent, visible participants with their own cryptographic identities, capable of observing conversations, contributing insights, and taking action right alongside people.

What Buzz Actually Changes About Workplace Chat

Buzz looks like a familiar group-chat workspace at first glance: communities, channels, threads, direct messages, file sharing, and voice. The difference sits inside those channels. AI agents are not a sidebar add-on or a @mention command. They have profiles, permissions, and a permanent presence in the spaces where work happens.

A product engineering channel might contain three human developers, a code-review agent watching repository events, a testing agent that posts build failures, a documentation agent drafting release notes, and an operations agent flagging anomalies—all operating in the same searchable conversation layer. The goal is not just faster answers. It is a shared record where request, context, action, output, review, and outcome stay connected.

Under the hood, Buzz is built on Nostr, a decentralized protocol based on cryptographically signed events and relay servers. Every participant—human or agent—gets a public key as identity and a private key to prove control. This is designed so messages, approvals, and workflow logs carry clear attribution: which agent performed an action, who operates it, and what permissions bound its behavior. The architecture supports self-hosted relays, giving organizations the option to run their own infrastructure rather than relying solely on Block’s hosted service, as described by Decrypt’s coverage.

Importantly, Buzz’s “decentralized” label does not mean messages freely roam between all relays. A message stays on the relay where it was sent, and relay operators control their own boundaries. That protects workplace confidentiality while preserving the portability and identity model that Nostr provides. The platform is model-agnostic: agents can be built with Claude Code, OpenAI Codex, Block’s own Goose framework, or other tools, giving teams freedom to choose their AI stack rather than being locked into one vendor.

What It Means for You

For Everyday Windows Users and Teams

If your team already uses Slack or Microsoft Teams, the immediate impact is small. However, Buzz is immediately available as a downloadable Windows app at no cost, and you can create a community or join one to experiment. The concrete promise is a workspace where an AI agent can sit in your project channel, monitor a shared data source, summarize a long thread, or create a draft—without anyone having to invoke it manually. Early impressions from The Information suggest the agent visibility could reduce the fragmentation that comes from scattering decisions across chat, GitHub, Jira, and email. The trade-off, of course, is that an agent that participates badly can create more noise than value, and organizations will need to define clear boundaries.

For IT Administrators and Security Teams

A Windows build exists, but enterprise deployment maturity is still early. There are no published MSI packages, Intune integration, or silent update controls as of this launch. The open-source code base does mean a technically capable team could audit and customize the client, but key management introduces a serious operational burden. If an employee loses a private key, identity recovery is not a simple password reset; if a key is compromised, impersonation is possible. Any IT department considering a trial should start by reviewing the security model, relay operation, and data-handling documentation Block has published, then test in an isolated environment with non-sensitive data. For Windows endpoints, be aware that the client currently requires manual installation and lacks group policy or registry-based configuration options.

For Developers and Open-Source Communities

Buzz is a natural playground. Since agents are first-class entities, developers can build custom agents that monitor internal APIs, push commit status, triage issues, or orchestrate simple workflows directly inside a team channel. The Nostr foundation also opens doors for cross-community identity; a developer could, in theory, carry the same cryptographic identity across different Buzz relays—something not possible in typical SaaS collaboration tools. The self-hosted relay option is especially attractive for open-source projects that want full control over their infrastructure and no vendor lock-in.

For Organizations Weighing Collaboration Tools

Buzz is not an out-of-the-box replacement for Microsoft Teams where compliance, eDiscovery, and Office integration are mandatory. But it challenges the underlying premise of tools that bolt AI onto human-only chat. If your organization is already experimenting with autonomous agents in development, security operations, or support, Buzz offers a more coherent model: agents with identities, logs, and visible accountability. The question is whether that structural advantage outweighs the enormous switching cost of migrating a team’s entire communication history, habits, and integrations.

How We Got Here: The Road to Agent-First Collaboration

Workplace chat has evolved incrementally. IRC gave way to Slack, which was acquired by Salesforce for $27.7 billion in 2021. Microsoft Teams bundled chat into the Office 365 ecosystem, and both platforms later added AI assistants. But those assistants remained largely in the side panel—search, summarize, rewrite. Meanwhile, the agent revolution accelerated. OpenAI’s GPT-4 and Anthropic’s Claude demonstrated that large language models could not just answer questions but execute multi-step tasks when connected to tools. AI coding agents became common; Devin and others showed agents could autonomously fix bugs and open pull requests.

Jack Dorsey has a history of betting on protocol-level change. Twitter’s acquisition and his early work on Bluesky reflected a belief that identity and data should be portable, not owned by a single platform. Block’s ventures in payments and Square cemented that philosophy around open financial systems. Buzz is an extension of that thinking into workplace software: give users cryptographic identities, let them host their own relays, and allow AI agents to join naturally rather than as external call-outs. The launch comes at a moment when enterprises are breaking away from single-provider AI lock-in and searching for model-agnostic, transparent AI integration.

The timing also exposes a pain point: chat platforms are sticky. Slack and Teams hold years of institutional memory, integrations, and workflow automations. Moving a team means rebuilding that social operating system, and Buzz’s agent-first pitch must prove that the efficiency gains justify the migration. So far, no collaboration tool has successfully dethroned a deeply embedded incumbent by offering a cleaner architecture alone.

What to Do Now: Practical First Steps for Windows Users

  1. Try the Buzz download on a secondary Windows device. Because the agent interaction model is unfamiliar, run it first outside your primary work machine. Create a small community with a few colleagues and a simple agent (Block’s Goose project offers a template) that monitors a low-risk data source. Observe how the agent behaves in a channel: Is it noisy? Does it produce useful output? How does the team react?

  2. Audit the security and key management surface. For any self-hosted relay, document exactly where private keys are generated and stored, how they are backed up, and what an offboarding procedure looks like if an employee departs. Without a corporate SSO integration, identity management is manual and risk-prone. Organizations should not treat agent credentials any differently from human user accounts—least-privilege principles must apply.

  3. Define an agent governance policy before going beyond a trial. Decide which channels agents can join, what data they may observe, whether their output must be reviewed before action, and which model providers are permitted. The Financial Express noted that Buzz’s open nature means a poorly configured agent could accidentally expose sensitive context to an external model endpoint. Start with the safest use case: a read-only agent that summarizes a public channel and posts a draft for a human to approve.

  4. Stay informed on Buzz’s enterprise roadmap. As of launch, Block has not announced Windows-specific enterprise tooling, but given the interest, IT managers should watch for silent installers, Group Policy support, and relay compliance certifications. The Yahoo Finance coverage speculates that self-hosted enterprise relays may arrive with administrative dashboards later in 2026, but no dates are confirmed.

  5. Keep your current collaboration tools. Buzz is a promising experiment, not yet a production replacement. Use it alongside Slack or Teams for a low-stakes project. The real value for most teams today is as a sandbox for learning how to design, constrain, and collaborate with persistent AI agents—lessons that will apply regardless of which platform ultimately dominates.

What to Watch Next

Buzz’s success depends on two things: execution quality and community adoption. Early iterations will be judged on whether agents truly reduce noise and improve throughput or simply turn channels into chaotic AI streams. Watch for developers building useful, well-behaved agents that solve narrow, measurable problems—those examples will influence broader team acceptance. Also monitor whether Block releases an enterprise relay that satisfies compliance teams; without it, large organizations will remain cautious. In the background, Slack and Microsoft are already accelerating their own agent platforms. Buzz’s biggest contribution may be forcing the entire collaboration industry to grapple with the governance, visibility, and accountability of AI in the workplace—a debate that every Windows team using AI agents will eventually have to join.