Here’s the honest rundown. Buzz is Jack Dorsey’s new free app that turns your AI agents into a full team of employees — talking to each other and working together inside one simple group chat that looks like Slack.
I built a whole working system in about 20 minutes: one agent writing blog posts, another designing images, a third checking their work. Here’s everything Buzz is and how to set it up.
Last updated: July 2026.
Key takeaways
- Buzz is Jack Dorsey’s free, open-source ‘Slack for AI agents’ — agents work together in channels.
- Plug in Claude Code, Codex or Goose (Goose is also from Block, Jack Dorsey’s company) — you only pay for models you already have.
- Build custom agents and agent teams — mine wrote, QC’d and published a full blog post autonomously.
- Agents inherit your existing Claude setup: skills, APIs, even your Obsidian vault.
- Learn to build agent teams like this in the AI Profit Boardroom.
What Is Buzz From Jack Dorsey?
Buzz is a free, open-source app from Jack Dorsey that works like Slack for AI agents: channels, threads, group chats — but the ‘teammates’ are AI agents that talk to each other and do real work. Type one message and a whole team springs into action.
It’s far more intuitive than most agent tools — if you’ve used Slack, you already understand it. Install the desktop app (easier than the GitHub route), sign in, and install your CLIs: Claude Code, Codex, and Goose — Goose being Block’s own agent, the same Jack Dorsey company behind Buzz.
The 20-Minute Team I Built
To show what it can genuinely do, here’s what I set up in about 20 minutes:
- A second-brain channel — an agent reads my Obsidian vault and suggests personalised automation ideas (a QC pipeline, SEO indexing, an email funnel for my community).
- An image designer agent — I gave it a logo and an example thumbnail I like; it produced genuinely nice variations.
- An SEO agent — it used my existing SEO skill and Google Search Console data (already wired into Claude), found new keywords, wrote a full post and published it straight to my WordPress — no logins, no new API keys.
- A quality controller — checks the team’s work and sends feedback before anything ships.
Agent Teams: The Killer Feature
You can group agents into teams with shared instructions — my ‘marketing team’ is the image designer, SEO agent and quality controller. Tag the whole team once and they coordinate autonomously: the SEO agent researched and drafted, called in the image designer when it needed visuals, and the QC agent reviewed everything.
The QC exchange was the standout: it caught capitalisation errors and called out that the claimed word count (1,970) was actually 1,580. The SEO agent fixed it, the QC confirmed “ship ready”, and the post went live — all from one message from me.
Why Buzz Is Different
| Feature | What it means |
|---|---|
| Uses your existing setup | Agents inherit Claude’s skills, APIs, GSC data, Obsidian — no re-wiring |
| Per-agent permissions | Each agent has its own keys, channels and audit trail — siloed access |
| Searchable history | Agents can search 6 months of history and post to threads |
| Invite humans | Teammates join and bring their own agents, with limited access |
| Activity view | Watch what each agent is doing in the background, with a reply timer |
| Free & open source | You only pay for model subscriptions you already have |
The Honest Caveats
Full transparency from testing: big tasks sometimes seem to time out — the agent finishes the work but you occasionally need to nudge it (“where are the images?”) for the reply. Slower models (like Opus 5 on heavy jobs) make this more noticeable. And it’s only about a week old, so expect rough edges and experimental features (workflows, polls, form channels, mobile) still in beta.
Buzz + Your Other Agents
Buzz plays well with the wider stack. I connect Hermes to it too — three different ways, covered in my Hermes + Buzz guide — and the workspace basics are in my free AI workspace guide and Buzz vs Slack comparison. You can even share your machine so community members run agents on remote models like Gemma 4.
Get the Full Training
The complete Buzz setup — custom agents, teams, and the workflows that publish content end-to-end — is inside the AI Profit Boardroom, with a full Buzz masterclass and daily updates.
New here? Start free with my AI Money Lab community (free AI course + 1,000+ AI agents), or grab a free strategy session.
FAQ
What is Buzz from Jack Dorsey?
A free, open-source app that works like Slack for AI agents — agents collaborate in channels, form teams, and do real work like writing, designing and publishing content.
Is Buzz free?
Yes — free and open source. You only pay for the model subscriptions you plug in (Claude Code, Codex), which you may already have.
What agents work with Buzz?
Claude Code, Codex and Goose (Block’s own agent) install directly; Hermes can connect too via desktop, relay or gateway.
Can Buzz agents really publish content?
Yes — my SEO agent found keywords from my Search Console data, wrote a post, had it quality-checked by another agent, and published it to WordPress autonomously.
What are the downsides?
It’s about a week old: big tasks can appear to time out (nudge for the reply), slower models lag, and several features are still experimental.
The Bottom Line
One message, a whole AI team in motion — that’s Buzz. The masterclass is in the AI Profit Boardroom.