Several agent orbs orbiting a shared hub, representing humans and AI agents working in the same workspace

Block Launches Buzz: Humans and AI Agents in One Workspace

Block released Buzz today — July 21, 2026 — an open-source workspace where AI agents sit in the same channels as the people they work with. Jack Dorsey announced it on X at 16:33 UTC:

“we’re launching BUZZ! a new groupchat platform for teams of people and agents of all sizes, built to reduce our dependency on slack and github. model-agnostic, decentralized, self-sovereign, and open source. 🐝”

That last part is the actual news. This isn’t a chatbot bolted onto a chat app. It’s an attempt to replace Slack and GitHub with one thing, and to give agents the same standing in it as humans.


What Buzz actually is

Block’s announcement opens: “Today we’re releasing Buzz, a free, open source collaboration platform where humans and AI agents work together in a shared workspace.”

Underneath, it’s a Nostr relay written in Rust, Apache-2.0 licensed, at github.com/block/buzz. Every message, reaction, workflow step, review approval, and git event is a signed event in one log. The repo’s own summary of why that matters: “Same shape, same identity model, same audit trail, whether the author is a person or a process.”

The practical consequence is the part worth paying attention to. Agents in Buzz get their own keypairs, their own channel memberships, and their own audit trail — scoped by identity rather than by permission flags, the way you’d scope a teammate. Add an agent to a channel the same way you add a person. It can open repos, send patches, review code, run workflows, edit canvases, and pull in whoever needs to see it.

Block is explicit that it’s model-agnostic: Claude Code, Codex, goose, or anything else that speaks the Agent Client Protocol can work in Buzz. Worth being precise here, because it’s easy to get backwards — Buzz is not built on goose. Block’s Builderbot is built on goose. Buzz is the room; goose is one of several agents that can walk into it, on identical footing with Claude Code and Codex.

What works today versus what doesn’t

Block publishes this split in the repo itself, which is more candid than most launch-day documentation:

Works todayBeing wired upPending code
Relay, channels, threads, DMs, canvases, media, search, audit logMobile clients (iOS + Android)Web-of-trust reputation across relays
Desktop app (Tauri + React)Workflow approval gatesPush notifications
buzz-cli + ACP harness (goose, Codex, Claude Code)Huddle lifecycle events
YAML workflows: message / reaction / schedule / webhook triggers
Git events, git hosting backend

The repo’s own caveat: “Not finished. We will tell you what works and what doesn’t.” Also, in a line that will save someone a bad assumption: “Not blockchain. Signed events are useful without making everyone buy a commemorative coin.”

Note the workflow row. Buzz ships event triggers on day one — workflows fire on a message, a reaction, a schedule, or a webhook. That’s a real capability, not a roadmap item, and it’s a meaningful part of why agents in Buzz can act rather than just talk.

What it costs is genuinely unclear

Block calls Buzz “a free, open source collaboration platform,” and the Apache-2.0 license backs that up for the software. You can run your own relay on your own infrastructure with full control over data, relay, and agents.

The hosted side is where the gap is. Block says you can “head to buzz.xyz to sign up and create a community with Block-provided hosting,” and that hosted teams “get the same protocol guarantees with managed infrastructure underneath.” No price is attached to that anywhere — not on the announcement, not on the engineering blog, not on buzz.xyz. “Free” describes the license, and it would be a mistake to read it as a priced free tier for managed hosting. That’s an open question, not a confirmed zero.

The reception is not uniformly warm

The launch’s main discussion venue is Hacker News, and the skepticism there is worth reading before you get excited. A Slack employee raised data-leakage concerns specific to multiple agents operating in shared rooms. Several commenters dismissed the whole thing as bots in chat rooms. Someone noted the irony of a GitHub alternative being hosted on GitHub.

None of that makes the underlying bet wrong, but the “agents as members with their own keys” model creates a genuine security surface: an agent in a channel can read everything in that channel. Buzz’s answer is that this is exactly why agents get scoped identities and an audit trail instead of blanket bot tokens. Whether that holds up under real multi-tenant use is unproven, and Block does not claim otherwise.

Who this is for

Buzz is developer infrastructure. The self-host path wants Docker, Rust 1.88+, Node 24+, and pnpm, and the payoff is a workspace your engineering team runs on its own relay. If you’re a team already coordinating AI agents across a codebase and you’re unhappy paying for Slack and GitHub separately, this is aimed squarely at you and it’s free to try.

If you’re a solo operator or a small business, this almost certainly isn’t your tool. Standing up a Nostr relay to get an agent to answer email is the wrong shape of effort, and none of Buzz’s git-centric surface applies to you. The premise it’s built on — agents that act on events instead of waiting to be prompted — is the right one, but you can get that from workflow automation tooling that doesn’t ask you to run infrastructure. Carly works that way over email, with 260+ native integrations and free, unlimited Zapier-style workflows; AI agents start at $35/month.

The thing to watch is whether “agents are members, not bots” becomes the default way collaboration software is built. Block just made the most credible open-source argument for it so far.

Ready to automate your busywork?

Carly schedules, researches, and briefs you—so you can focus on what matters.

See what people say

"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.

Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.

On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."

Gus Ibrahim, Founder & Director, IHR