# Jack Dorsey's Buzz App Lets AI Agents Join Your Team Chats > Source: > Published: 2026-08-01 00:00:00+00:00 # Jack Dorsey's Buzz App Lets AI Agents Join Your Team Chats Jack Dorsey's Buzz app invites Claude Code, Codex, and other AI agents into group chats. Here's how it works and who it's for. ## What is Buzz, and why is Jack Dorsey building it? Buzz is a group chat app from Jack Dorsey’s Block, built on top of his open-source Goose AI agent framework. The pitch is simple: instead of a chat app where humans talk to each other and occasionally paste in an AI response, Buzz treats AI agents as first-class members of the conversation. You can invite Claude Code, Codex, or other coding and reasoning agents into a channel the same way you’d invite a coworker, and they participate directly, reading messages, responding, and acting on tasks in real time. Dorsey has spent the last few years pushing decentralized, protocol-based software (Bluesky’s AT Protocol being the best-known example), and Buzz follows that pattern rather than the closed, single-vendor model of Slack or Microsoft Teams. The goal is a chat layer where agents from different companies and different model providers can sit in the same room and work together, not just one company’s chatbot bolted onto a messaging product. ## How does it work in practice? ## Plans first. *Then code.* Remy writes the spec, manages the build, and ships the app. The core idea is that a Buzz channel isn’t just a thread of messages. It’s a shared workspace where each participant, human or agent, can see the same context and act on it. When you add an agent like Claude Code or Codex to a channel, it isn’t a separate sidebar assistant. It’s in the conversation, able to read what other agents and humans have posted, respond to direct questions, and carry out tasks like writing code, reviewing a document, or checking work. Because Buzz is built on Goose, an open agent framework rather than a single proprietary model, it’s designed to be model-agnostic. That means a team could run a coding-focused agent from one provider alongside a research or writing agent from another, in the same channel, without needing separate apps or separate logins for each. ## What can multiple agents in one chat actually do? The most interesting demonstrated use case is having agents critique each other’s work inside the same channel. Rather than a human bouncing a document or a piece of code between two separate AI tools and manually copying feedback back and forth, you can have one agent produce a draft (code, a plan, a piece of writing) and a second agent review it, flag issues, and suggest changes, all visible in the same thread. This turns a chat channel into something closer to a lightweight multi-agent pipeline. A human can set up the task, invite two or three agents suited to different parts of it, and watch them hand work back and forth with the human stepping in only to steer or approve. It’s a different pattern from the now-familiar “one agent, one long autonomous task” setup (like letting a coding agent run for hours unsupervised). Here the value is in agents checking each other’s output in a shared, visible space, which cuts down on the blind spots any single model has when grading its own homework. ## How is this different from Slack or Teams with a bot bolted on? Slack, Teams, and similar tools have added AI features, but mostly as an assistant you summon into a thread or a bot account that posts on a schedule. The agent is a guest, not a participant with ongoing context across a project. Buzz flips that: agents are treated as members of the channel by default, with persistent access to the conversation history and the ability to initiate actions, not just respond when tagged. The other structural difference is decentralization. Slack and Teams are closed platforms controlled by one vendor. Buzz, built on the Goose framework, is designed to be open, so in theory a team isn’t locked into a single AI provider’s assistant. You can mix agents from different vendors in the same channel, which matters as more teams end up using several different coding and reasoning models rather than standardizing on one. ## Is Buzz actually useful, or is it a novelty right now? It’s early. As with most decentralized-protocol projects, the value of Buzz depends heavily on adoption, on other teams and tool builders actually connecting their agents to it rather than staying inside walled-garden platforms. A single team experimenting with two or three agents critiquing each other’s code in a channel is a genuinely useful workflow today. Whether it becomes how teams collaborate at scale depends on whether more agent providers plug into the same open framework instead of pushing their own closed chat surfaces. It’s also worth putting Buzz in context with the broader trend it’s part of. Agentic features are showing up everywhere right now, not just in dedicated agent tools but inside existing platforms. Meta AI recently added the ability to connect calendar and email so its assistant can act on your schedule directly inside a chat interface. Google added a “create image” button inside Google Earth that runs Nano Banana against a 3D map view so you can generate a reimagined version of a real place with a text prompt, free to use, no separate app required. The common thread is that AI is moving from a separate destination you visit into a layer embedded in tools you already use. Buzz is doing the same thing for team chat, just with a decentralized, multi-agent twist instead of a single vendor’s assistant. For teams already leaning on multiple coding agents (Claude Code, Codex, and others) day to day, a shared channel where those agents can compare notes without a human relaying messages between separate tabs is a real efficiency gain. For teams happy with a single-vendor AI assistant inside their existing chat tool, Buzz is more of a “watch this space” than an immediate switch. ## Frequently Asked Questions ### What is Jack Dorsey’s Buzz app? Buzz is a group chat application from Block, built on the open-source Goose agent framework, that lets AI agents like Claude Code and Codex join channels as full participants alongside humans, rather than as bolted-on assistants. ### Is Buzz a Slack competitor? It competes for the same team-chat use case, but its core difference is treating AI agents as first-class channel members with persistent context, and being built on an open, model-agnostic framework instead of a single closed platform. ### What is Goose, and how does it relate to Buzz? Goose is Block’s open-source AI agent framework. Buzz is built on top of it, which is why Buzz can support agents from different providers in the same chat rather than locking users into one company’s model. ### Can AI agents in Buzz review each other’s work? Yes. One demonstrated pattern has one agent produce output, like code or a draft, while a second agent in the same channel critiques or checks that work, with the exchange visible to any humans in the channel. ### Do I need to use a specific AI model to use Buzz? No. Because it’s built on the open Goose framework, Buzz is designed to work with multiple agents and models rather than requiring one specific provider’s assistant.